oBIX 1.0

Committee Specification 01, 5 December 2006

Document identifier:

obix-1.0-cs-01

Location:

http://www.oasis-open.org/committees/obix

Technical Committee:

OASIS Open Building Information Exchange TC

Chairs:

Toby Considine, University of North Carolina Chapel Hill

Paul Ehrlich, Building Intelligence Group

Editor:

Brian Frank, Tridium

Abstract:

This document specifies an object model and XML format used for machine-to-machine (M2M) communication.

Status:

This document was last revised or approved by the Open Building Information Exchange TC on the above date. The level of approval is also listed above. Check the current location noted above for possible later revisions of this document. This document is updated periodically on no particular schedule.

Technical Committee members should send comments on this specification to the Technical Committee’s email list. Others should send comments to the Technical Committee by using the “Send A Comment” button on the Technical Committee’s web page at www.oasis-open.org/committees/obix.

For information on whether any patents have been disclosed that may be essential to implementing this specification, and any offers of patent licensing terms, please refer to the Intellectual Property Rights section of the Technical Committee web page (www.oasis-open.org/committees/obix/ipr.php).

The non-normative errata page for this specification is located at www.oasis-open.org/committees/obix.

Notices

OASIS takes no position regarding the validity or scope of any intellectual property or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; neither does it represent that it has made any effort to identify any such rights. Information on OASIS's procedures with respect to rights in OASIS specifications can be found at the OASIS website. Copies of claims of rights made available for publication and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification, can be obtained from the OASIS Executive Director.

OASIS invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights which may cover technology that may be required to implement this specification. Please address the information to the OASIS Executive Director.

This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself does not be modified in any way, such as by removing the copyright notice or references to OASIS, except as needed for the purpose of developing OASIS specifications, in which case the procedures for copyrights defined in the OASIS Intellectual Property Rights document must be followed, or as required to translate it into languages other than English.

The limited permissions granted above are perpetual and will not be revoked by OASIS or its successors or assigns.

This document and the information contained herein is provided on an “AS IS” basis and OASIS DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Table of Contents

3.6 Extendibility. 13

4.16.1 displayName. 17

5.3 HTTP Relative URIs. 21

5.4 Fragment URIs. 21

6 Contracts. 22

6.1 Contract Terminology. 22

6.2 Contract List 22

6.3 Is Attribute. 23

6.4 Contract Inheritance. 23

6.5 Override Rules. 24

6.6 Multiple Inheritance. 25

6.6.1 Flattening. 25

6.6.2 Mixins. 25

6.7 Contract Compatibility. 26

6.8 Lists (and Feeds) 26

7 XML. 28

7.1 Design Philosophy. 28

7.2 XML Syntax 28

7.3 XML Encoding. 28

7.4 XML Decoding. 29

7.5 XML Namespace. 29

7.6 Namespace Prefixes in Contract Lists. 29

8 Operations. 30

9 Object Composition. 31

10.1 Request / Response. 33

11 Core Contract Library. 38

12.1 WatchService. 41

12.2 Watch. 41

12.2.1 Watch.add. 42

12.2.2 Watch.remove. 43

12.2.3 Watch.pollChanges. 43

12.2.4 Watch.pollRefresh. 43

12.2.5 Watch.lease. 43

12.2.6 Watch.delete. 44

12.3 Watch Depth. 44

12.4 Feeds. 44

13 Points. 46

13.1 Writable Points. 46

14 History. 47

14.1 History Object 47

14.2 History Queries. 47

14.2.1 HistoryFilter 47

14.2.2 HistoryQueryOut 48

14.2.3 HistoryRecord. 48

14.2.4 History Query Example. 48

14.3 History Rollups. 49

14.3.1 HistoryRollupIn. 49

14.3.2 HistoryRollupOut 49

14.3.3 HistoryRollupRecord. 49

14.3.4 Rollup Calculation. 49

14.4 History Feeds. 51

15 Alarming. 52

15.1 Alarm States. 52

15.1.1 Alarm Source. 52

15.1.2 StatefulAlarm and AckAlarm.. 52

15.2 Alarm Contracts. 53

15.2.1 Alarm.. 53

15.2.2 StatefulAlarm.. 53

15.2.3 AckAlarm.. 53

15.2.4 PointAlarms. 54

15.3 AlarmSubject 54

15.4 Alarm Feed Example. 54

16 Security. 56

16.1 Error Handling. 56

16.2 Permission based Degradation. 56

17 HTTP Binding. 57

17.1 Requests. 57

17.2 Security. 57

17.3 Localization. 57

18 SOAP Binding. 59

18.1 SOAP Example. 59

18.2 Error Handling. 59

18.3 Security. 59

18.4 Localization. 59

18.5 WSDL. 60

Appendix A. Revision History. 62

1 Overview

oBIX is designed to provide access to the embedded software systems which sense and control the world around us. Historically integrating to these systems required custom low level protocols, often custom physical network interfaces. But now the rapid increase in ubiquitous networking and the availability of powerful microprocessors for low cost embedded devices is weaving these systems into the very fabric of the Internet. Generically the term M2M for Machine-to-Machine describes the transformation occurring in this space because it opens a new chapter in the development of the Web - machines autonomously communicating with each other. The oBIX specification lays the groundwork building this M2M Web using standard, enterprise friendly technologies like XML, HTTP, and URIs.

The following design points illustrate the problem space oBIX attempts to solve:

XML: representing M2M information in a standard XML syntax;
Networking: transferring M2M information in XML over the network;
Normalization: standard representations for common M2M features: points, histories, and alarms;
Foundation: providing a common kernel for new standards;

1.1 XML

The principle requirement of oBIX is to develop a common XML syntax for representing information from diverse M2M systems. The design philosophy of oBIX is based on a small, but extensible data model which maps to a simple fixed XML syntax. This core object model and it’s XML syntax is simple enough to capture entirely in one illustration provided in Chapter 4. The object model’s extensibility allows for the definition of new abstractions through a concept called contracts. The majority of the oBIX specification is actually defined in oBIX itself through contracts.

1.2 Networking

Once we have a way to represent M2M information in XML, the next step is to provide standard mechanisms to transfer it over networks for publication and consumption. oBIX breaks networking into two pieces: an abstract request/response model and a series of protocol bindings which implement that model. Version 1.0 of oBIX defines two protocol bindings designed to leverage existing web service infrastructure: an HTTP REST binding and a SOAP binding.

1.3 Normalization

There are a few concepts which have broad applicability in systems which sense and control the physical world. Version 1.0 of oBIX provides a normalized representation for three of these:

Points: representing a single scalar value and it’s status – typically these map to sensors, actuators, or configuration variables like a setpoint;
Histories: modeling and querying of time sampled point data. Typically edge devices collect a time stamped history of point values which can be fed into higher level applications for analysis;
Alarming: modeling, routing, and acknowledgment of alarms. Alarms indicate a condition which requires notification of either a user or another application.

1.4 Foundation

The requirements and vertical problem domains for M2M systems are immensely broad – too broad to cover in one single specification. oBIX is deliberately designed as a fairly low level specification, but with a powerful extension mechanism based on contracts. The goal of oBIX is to lay the groundwork for a common object model and XML syntax which serves as the foundation for new specifications. It is hoped that a stack of specifications for vertical domains can be built upon oBIX as a common foundation.

2 Quick Start

This chapter is for those eager beavers who want to immediately jump right into oBIX and all its angle bracket glory. The best way to begin is to take a simple example that anybody is familiar with – the staid thermostat. Let’s assume we have a very simple thermostat. It has a temperature sensor which reports the current space temperature and it has a setpoint that stores the desired temperature. Let’s assume our thermostat only supports a heating mode, so it has a variable that reports if the furnace should currently be on. Let’s take a look at what our thermostat might look like in oBIX XML:

</obj>

The first thing to notice is that there are three element types. In oBIX there is a one-to-one mapping between objects and elements. Objects are the fundamental abstraction used by the oBIX data model. Elements are how those objects are expressed in XML syntax. This document uses the term object and sub-objects, although you can substitute the term element and sub-element when talking about the XML representation.

The root obj element models the entire thermostat. Its href attribute identifies the URI for this oBIX document. There are three child objects for each of the thermostat’s variables. The real objects store our two floating point values: space temperature and setpoint. The bool object stores a boolean variable for furnace state. Each sub-element contains a name attribute which defines the role within the parent. Each sub-element also contains a val attribute for the current value. Lastly we see that we have annotated the temperatures with an attribute called units so we know they are in Fahrenheit, not Celsius (which would be one hot room). The oBIX specification defines a bunch of these annotations which are called facets.

In real life, sensor and actuator variables (called points) imply more semantics than a simple scalar value. In other cases such as alarms, it is desirable to standardize a complex data structure. oBIX captures these concepts into contracts. Contracts allow us to tag objects with normalized semantics and structure.

Let’s suppose our thermostat’s sensor is reading a value of -412°F? Clearly our thermostat is busted, so we should report a fault condition. Let’s rewrite the XML to include the status facet and to provide additional semantics using contracts:

<real name="spaceTemp" is="obix:Point"

val="-412.0" status="fault"

units="obix:units/fahrenheit"/>

<real name="setpoint" is="obix:Point"

val="72.0"

unit="obix:units/fahrenheit"/>

</obj>

Notice that each of our three scalar values are tagged as obix:Points via the is attribute. This is a standard contract defined by oBIX for representing normalized point information. By implementing these contracts, clients immediately know to semantically treat these objects as points.

Contracts play a pivotal role in oBIX for building new abstractions upon the core object model. Contracts are slick because they are just normal objects defined using standard oBIX syntax (see Chapter 13 to take sneak peak the point contracts).

3 Architecture

The oBIX architecture is based on the following principles:

Object Model: a concise object model used to define all oBIX information.
XML Syntax: a simple XML syntax for expressing the object model.
URIs: URIs are used to identify information within the object model.
REST: a small set of verbs is used to access objects via their URIs and transfer their state via XML.
Contracts: a template model for expressing new oBIX “types”.
Extendibility: providing for consistent extendibility using only these concepts.

3.1 Object Model

All information in oBIX is represented using a small, fixed set of primitives. The base abstraction for these primitives is cleverly called object. An object can be assigned a URI and all objects can contain other objects.

There are eight special kinds of value objects used to store a piece of simple information:

bool: stores a boolean value - true or false;
int: stores an integer value;
real: stores a floating point value;
str: stores a UNICODE string;
enum: stores an enumerated value within a fixed range;
abstime: stores an absolute time value (timestamp);
reltime: stores a relative time value (duration or time span);
uri: stores a Universal Resource Identifier;

Note that any value object can also contain sub-objects. There are also a couple of other special object types: list, op, feed, ref and err.

3.2 XML

oBIX is all about a simple XML syntax to represent its underlying object model. Each of the object types map to one type of element. The value objects represent their data value using the val attribute. All other aggregation is simply nesting of elements. A simple example to illustrate:

</obj>

</list>

</obj>

Note in this simple example how the href attribute specifies URI references which may be used to fetch more information about the object. Names and hrefs are discussed in detail in Chapter 5.

3.3 URIs

No architecture is complete without some sort of naming system. In oBIX everything is an object, so we need a way to name objects. Since oBIX is really about making information available over the web using XML, it makes to sense to leverage the venerable URI (Uniform Resource Identifier). URIs are the standard way to identify “resources” on the web.

Often URIs also provide information about how to fetch their resource - that’s why they are often called URLs (Uniform Resource Locator). From a practical perspective if a vendor uses HTTP URIs to identify their objects, you can most likely just do a simple HTTP GET to fetch the oBIX document for that object. But technically, fetching the contents of a URI is a protocol binding issue discussed in later chapters.

The value of URIs are that they come with all sorts of nifty rules already defined for us (see RFC 3986). For example URIs define which characters are legal and which are illegal. Of great value to oBIX is URI references which define a standard way to express and normalize relative URIs. Plus most programming environments have libraries to manage URIs so developers don’t have to worry about nitty gritty normalization details.

3.4 REST

Many savvy readers may be thinking that objects identified with URIs and passed around as XML documents is starting to sound a lot like REST – and you would be correct. REST stands for REpresentational State Transfer and is an architectural style for web services that mimics how the World Wide Web works. The WWW is basically a big web of HTML documents all hyperlinked together using URIs. Likewise, oBIX is basically a big web of XML object documents hyperlinked together using URIs.

REST is really more of a design style, than a specification. REST is resource centric as opposed to method centric - resources being oBIX objects. The methods actually used tend to be a very small fixed set of verbs used to work generically with all resources. In oBIX all network requests boil down to three request types:

Read: an object
Write: an object
Invoke: an operation

3.5 Contracts

In every software domain, patterns start to emerge where many different object instances share common characteristics. For example in most systems that model people, each person probably has a name, address, and phone number. In vertical domains we may attach domain specific information to each person. For example an access control system might associate a badge number with each person.

In object oriented systems we capture these patterns into classes. In relational databases we map them into tables with typed columns. In oBIX we capture these patterns using a concept called contracts, which are standard oBIX objects used as a template. Contracts are more nimble and flexible than strongly typed schema languages, without the overhead of introducing new syntax. A contract document is parsed just like any other oBIX document. In geek speak contracts are a combination of prototype based inheritance and mixins.

Why do we care about trying to capture these patterns? The most important use of contracts is by the oBIX specification itself to define new standard abstractions. It is just as important for everyone to agree on normalized semantics as it is as on syntax. Contracts also provide the definitions needed to map to the OO guy’s classes or the relational database guy’s tables.

3.6 Extendibility

We want to use oBIX as a foundation for developing new abstractions in vertical domains. We also want to provide extendibility for vendors who implement oBIX across legacy systems and new product lines. Additionally, it is common for a device to ship as a blank slate and be completely programmed in the field. This leaves us with a mix of standards based, vendor based, and even project based extensions.

The principle behind oBIX extendibility is that anything new is defined strictly in terms of objects, URIs, and contracts. To put it another way - new abstractions don’t introduce any new XML syntax or functionality that client code is forced to care about. New abstractions are always modeled as standard trees of oBIX objects, just with different semantics. That doesn’t mean that higher level application code never changes to deal with new abstractions, but the core stack that deals with networking and parsing shouldn’t have to change.

This extendibility model is similar to most mainstream programming languages such as Java or C#. The syntax of the core language is fixed with a built in mechanism to define new abstractions. Extendibility is achieved by defining new class libraries using the language’s fixed syntax. This means I don’t have to update the compiler every time some one adds a new class.

4 Object Model

The oBIX specification is based on a small, fixed set of object types. These object types map one to one to an XML element type. The oBIX object model is summarized in the following illustration. Each box represents a specific object type (and XML element name). Each object type also lists its supported attributes.

4.1 obj

The root abstraction in oBIX is object, modeled in XML via the obj element. Every XML element in oBIX is a derivative of the obj element. Any obj element or its derivatives can contain other obj elements. The attributes supported on the obj element include:

name: defines the object’s purpose in its parent object (discussed in the Chapter 5);
href: provides a URI reference for identifying the object (discussed in the Chapter 5);
is: defines the contracts the object implements (discussed in Chapter 6);
null: support for null objects (discussed in Section 4.15)
facets: a set of attributes used to provide meta-data about the object (discussed in Section 4.16);
val: an attribute used only with value objects (bool, int, real, str, enum, abstime, reltime, and uri) to store the actual value;

The contract definition of obj:

4.2 bool

The bool object represents a boolean condition of either true or false. Its val attribute maps to xs:boolean defaulting to false. The literal value of a bool must be “true” or “false” (the literals “1” and “0” are not allowed). The contract definition:

An example:

4.3 int

The int type represents an integer number. Its val attribute maps to xs:long as a 64-bit integer with a default of 0. The contract definition:

An example:

4.4 real

The real type represents a floating point number. Its val attribute maps to xs:double as a IEEE 64-bit floating point number with a default of 0. The contract definition:

An example:

4.5 str

The str type represents a string of Unicode characters. Its val attribute maps to xs:string with a default of the empty string. The contract definition:

An example:

4.6 enum

The enum type is used to represent a value which must match a finite set of values. The finite value set is called the range. The val attribute of an enum is represented as a string key using xs:string. Enums default to null. The range of an enum is declared via facets using the range attribute. The contract definition:

An example:

4.7 abstime

The abstime type is used to represent an absolute point in time. Its val attribute maps to xs:dateTime. Abstimes default to null. The contract definition:

An example for 9 March 2005 at 1:30PM GMT:

4.8 reltime

The reltime type is used to represent a relative duration of time. Its val attribute maps to xs:duration with a default of 0sec. The contract definition:

An example of 15 seconds:

4.9 uri

The uri type is used to store a URI reference. Unlike a plain old str, a uri has a restricted lexical space as defined by RFC 3986 and XML Schema anyURI type. Most URIs will also be a URL, meaning that they identify a resource and how to retrieve it (typically via HTTP). The contract:

An example for the oBIX home page:

4.10 list

The list object is a specialized object type for storing a list of other objects. The primary advantage of using a list versus a generic obj is that lists can specify a common contract for their contents using the of attribute. If specified the of attribute must be a list of URIs formatted as a contract list. The definition of list is:

An example list of strings:

</list>

Lists are discussed in greater detail along with contracts in section 6.8.

4.11 ref

The ref object is used to create an out of document reference to another oBIX object. It the oBIX equivalent of the HTML anchor tag. The contract definition:

A ref element must always specify a href attribute. References are discussed in detail in section 9.2.

4.12 err

The err object is a special object used to indicate an error. Its actual semantics are context dependent. Typically err objects should include a human readable description of the problem via the display attribute. The contract definition:

4.13 op

The op object is used to define an operation. All operations take one input object as a parameter, and return one object as an output. The input and output contracts are defined via the in and out attributes. The contract definition:

Operations are discussed in detail in Chapter 8.

4.14 feed

The feed object is used to define a topic for a feed of events. Feeds are used with watches to subscribe to a stream of events such as alarms. A feed should specify the event type it fires via the of attribute. The in attribute can be used to pass an input argument when subscribing to the feed (a filter for example).

Feeds are subscribed via Watches discussed in Chapter 12.

4.15 Null

All objects support the concept of null. Null is the absence of a value. Null is indicated using the null attribute with a boolean value. All objects default null to false with the exception of enum and abstime (since any other default would be confusing).

Null is inherited from contracts a little differently than other attributes. See Section 6.4 for details.

4.16 Facets

All objects can be annotated with a predefined set of attributes called facets. Facets provide additional meta-data about the object. The set of available facets is: displayName, display, icon, min, max, precision, range, and unit. Vendors which wish to annotate objects with additional facets should consider using XML namespace qualified attributes.

4.16.1 displayName

The displayName facet provides a localized human readable name of the object stored as a xs:string:

Typically the displayName facet should be a localized form of the name attribute. There are no restrictions on displayName overrides from the contract (although it should be uncommon since displayName is just a human friendly version of name).

4.16.2 display

The display facet provides a localized human readable description of the object stored as a xs:string:

There are no restrictions on display overrides from the contract.

The display attribute serves the same purpose as Object.toString() in Java or C#. It provides a general way to specify a string representation for all objects. In the case of value objects (like bool or int) it should provide a localized, formatted representation of the val attribute.

4.16.3 icon

The icon facet provides a URI reference to a graphical icon which may be used to represent the object in an user agent:

The contents of the icon attribute must be a URI to an image file. The image file is preferably a 16x16 PNG file. There are no restrictions on icon overrides from the contract.

4.16.4 min

The min facet is used to define an inclusive minimum value:

The contents of the min attribute must match its associated val type. The min facet is used with int, real , abstime, and reltime to define an inclusive lower limit of the value space. It is used with str to indicate the minimum number of Unicode characters of the string. It is used with list to indicate the minimum number of child objects (named or unnamed). Overrides of the min facet may only narrow the value space using a larger value. The min facet must never be greater than the max facet (although they can be equal).

4.16.5 max

The max facet is used to define an inclusive maximum value:

The contents of the max attribute must match its associated val type. The max facet is used with int, real, abstime, and reltime to define an inclusive upper limit of the value space. It is used with str to indicate the maximum number of Unicode characters of the string. It is used with list to indicate the maximum number of child objects (named or unnamed). Overrides of the max facet may only narrow the value space using a smaller value. The max facet must never be less than the min facet (although they can be equal).

4.16.6 precision

The precision facet is used to describe the number of decimal places to use for a real value:

The contents of the precision attribute must be xs:int. The value of the precision attribute equates to the number of meaningful decimal places. In the example above, the value of 2 indicates two meaningful decimal places: “75.04”. Typically precision is used by client applications which do their own formatting of real values. There are no restrictions on precision overrides.

4.16.7 range

The range facet is used to define the value space of an enumeration. A range attribute is a URI reference to an obix:Range object (see section 11.2 for the definition). It is used with the bool and enum object types:

The override rule for range is that the specified range must inherit from the contract’s range. Enumerations are funny beasts in that specialization of an enum usually involves adding new items to the range. Technically this is widening the enum’s value space, rather than narrowing it. But in practice, adding items into the range is what we desire.

4.16.8 status

The status facet is used to annotate an object about the quality and state of the information:

Status is an enumerated string value with one of the following values (ordered by priority):

disabled: This state indicates that the object has been disabled from normal operation (out of service). In the case of operations and feeds, this state is used to disable support for the operation or feed.
fault: The fault state indicates that the data is invalid or unavailable due to a failure condition - data which is out of date, configuration problems, software failures, or hardware failures. Failures involving communications should use the down state.
down: The down state indicates a communication failure.
unackedAlarm: The unackedAlarm state indicates there is an existing alarm condition which has not been acknowledged by a user – it is the combination of the alarm and unacked states. The difference between alarm and unackedAlarm is that alarm implies that a user has already acknowledged the alarm or that no human acknowledgement is necessary for the alarm condition. The difference between unackedAlarm and unacked is that the object has returned to a normal state.
alarm: This state indicates the object is currently in the alarm state. The alarm state typically means that an object is operating outside of its normal boundaries. In the case of an analog point this might mean that the current value is either above or below its configured limits. Or it might mean that a digital sensor has transitioned to an undesired state. See Alarming (Chapter 15) for additional information.
unacked: The unacked state is used to indicate a past alarm condition which remains unacknowledged.
overridden: The overridden state means the data is ok, but that a local override is currently in effect. An example of an override might be the temporary override of a setpoint from it’s normal scheduled setpoint.
ok: The ok state indicates normal status. This is the assumed default state for all objects.

Status must be one of the enumerated strings above. It might be possible in the native system to exhibit multiple status states simultaneously, however when mapping to oBIX the highest priority status should be chosen – priorities are ranked from top (disabled) to bottom (ok).

4.16.9 unit

The unit facet defines a unit of measurement. A unit attribute is a URI reference to a obix:Unit object (see section 11.5 for the contract definition). It is used with the int and real object types:

It is recommended that the unit facet not be overridden if declared in a contract. If it is overridden, then the override should use a Unit object with the same dimensions as the contract (it must measure the same physical quantity).

4.16.10 writable

The writable facet specifies if this object can be written by the client. If false (the default), then the object is read-only. It is used with all objects except operations and feeds:

5 Naming

All oBIX objects have two potential identifiers: name and href. Name is used to define the role of an object within its parent. Names are programmatic identifiers only; the displayName facet should be used for human interaction. Naming convention is to use camel case with the first character in lowercase. The primary purpose of names is to attach semantics to sub-objects. Names are also used to indicate overrides from a contract. A good analogy to names is the field/method names of a class in Java or C#.

Hrefs are used to attach URIs to objects. An href is always a URI reference, which means it might be a relative URI that requires normalization against a base URI. The exception to this rule is the href of the root object in an oBIX document – this href must be an absolute URI, not a URI reference. This allows the root object’s href to be used as the effective base URI (xml:base) for normalization. A good analogy is hrefs in HTML or XLink.

Some objects may have both a name and an href, just a name, just an href, or neither. It is common for objects within a list to not use names, since most lists are unnamed sequences of objects. The oBIX specification makes a clear distinction between names and hrefs - you should not assume any relationship between names and hrefs. From a practical perspective many vendors will likely build an href structure that mimics the name structure, but client software should never assume such a relationship.

5.1 Name

The name of an object is represented using the name attribute. Names are programmatic identifiers with restrictions on their valid character set. A name must contain only ASCII letters, digits, underbar, or dollar signs. A digit may not be used as the first character. Convention is to use camel case with the first character in lower case: “foo”, “fooBar”, “thisIsOneLongName”. Within a given object, all of its direct children must have unique names. Objects which don’t have a name attribute are called unnamed objects. The root object of an oBIX document should not specify a name attribute (but almost always has an absolute href URI).

5.2 Href

The href of an object is represented using the href attribute. If specified, the root object must have an absolute URI. All other hrefs within an oBIX document are treated as URI references which may be relative. Because the root href is always an absolute URI, it may be used as the base for normalizing relative URIs within the document. The formal rules for URI syntax and normalization are defined in RFC 3986. We consider a few common cases that serve as design patterns within oBIX in Section 5.3.

As a general rule every object accessible for a read must specify a URI. An oBIX document returned from a read request must specify a root URI. However, there are certain cases where the object is transient, such as a computed object from an operation invocation. In these cases there may not be a root URI, meaning there is no way to retrieve this particular object again. If no root URI is provided, then the server’s authority URI is implied to be the base URI for resolving relative URI references.

5.3 HTTP Relative URIs

Vendors are free to use any URI scheme, although the recommendation is to use HTTP URIs since they have well defined normalization semantics. This section provides a summary of how HTTP URI normalization should work within oBIX client agents. The general rules are:

If the URI starts with “scheme:” then it is an globally absolute URI
If the URI starts with a single slash, then it is server absolute URI
If the URI starts with a “#”, then it is a fragment identifier (discussed in next section)
If the URI starts with “../”, then the path must backup from the base

Otherwise the URI is assumed to be a relative path from the base URI

Some examples:

http://server/a + http://overthere/x ® http://overthere/x

http://server/a + /x/y/z ® http://server/x/y/z

http://server/a/b + c ® http://server/a/c

http://server/a/b/ + c ® http://server/a/b/c

http://server/a/b + c/d ® http://server/a/c/d

http://server/a/b/ + c/d ® http://server/a/b/c/d

http://server/a/b + ../c ® http://server/c

http://server/a/b/ + ../c ® http://server/a/c

Perhaps one of the trickiest issues is whether the base URI ends with slash. If the base URI doesn’t end with a slash, then a relative URI is assumed to be relative to the base’s parent (to match HTML). If the base URI does end in a slash, then relative URIs can just be appended to the base. In practice, systems organized into hierarchical URIs should always specify the base URI with a trailing slash. Retrieval with and without the trailing slash should be supported with the resulting document always adding the implicit trailing slash in the root object’s href.

5.4 Fragment URIs

It is not uncommon to reference an object internal to an oBIX document. This is achieved using fragment URI references starting with the “#”. Let’s consider the example:

</list>

</obj>

In this example there are two objects with a range facet referencing a fragment URI. Any URI reference starting with “#” must be assumed to reference an object within the same oBIX document. Clients should not perform another URI retrieval to dereference the object. In this case the object being referenced is identified via the href attribute.

In the example above the object with an href of “onOff” is both the target of the fragment URI, but also has the absolute URI “http://server/whatever/onOff”. But suppose we had an object that was the target of a fragment URI within the document, but could not be directly addressed using an absolute URI? In that case the href attribute should be a fragment identifier itself. When an href attribute starts with “#” that means the only place it can be used is within the document itself:

…

…

6 Contracts

Contracts are a mechanism to harness the inherit patterns in modeling oBIX data sources. What is a contract? Well basically it is just a normal oBIX object. What makes a contract object special, is that other objects reference it as a “template object” using the is attribute.

So what does oBIX use contracts for? Contracts solve many problems in oBIX:

Semantics: contracts are used to define “types” within oBIX. This lets us collectively agree on common object definitions to provide consistent semantics across vendor implementations. For example the Alarm contract ensures that client software can extract normalized alarm information from any vendor’s system using the exact same object structure.
Defaults: contracts also provide a convenient mechanism to specify default values. For example the Alarm contract provides a specification for all the default values which don’t need to be passed over the network for every read.
Type Export: it is likely that many vendors will have a system built using a statically typed language like Java or C#. Contracts provide a standard mechanism to export type information in a format that all oBIX clients can consume.

Why use contracts versus other approaches? There are certainly lots of ways to solve the above problems. The benefit of the contract design is its flexibility and simplicity. Conceptually contracts provide an elegant model for solving many different problems with one abstraction. From a specification perspective, we can define new abstractions using the oBIX XML syntax itself. And from an implementation perspective, contracts give us a machine readable format that clients already know how to retrieve and parse – to use OO lingo, the exact same syntax is used to represent both a class and an instance.

6.1 Contract Terminology

In order to discuss contracts, it is useful to define a couple of terms:

Contract: is a reusable object definition expressed as a standard oBIX XML document. Contracts are the templates or prototypes used as the foundation of the oBIX type system.
Contract List: is a list of one or more URIs to contract objects. It is used as the value of the is, of, in and out attributes. The list of URIs is separated by the space character. You can think of a contract list as a type declaration.
Implements: when an object specifies a contract in its contract list, the object is said to implement the contract. This means that the object is inheriting both the structure and semantics of the specified contract.
Implementation: an object which implements a contract is said to be an implementation of that contract.

6.2 Contract List

The syntax of a contract list attribute is a list of URI references to other oBIX objects. It is used as the value of the is, of, in and out attributes. The URIs within the list are separated by the space character (Unicode 0x20). Just like the href attribute, a contract URI can be an absolute URI, server relative, or even a fragment reference. The URIs within a contract list may be scoped with an XML namespace prefix (see Section 7.6).

6.3 Is Attribute

An object defines the contracts it implements via the is attribute. The value of the is attribute is a contract list. If the is attribute is unspecified, then the following rules are used to determine the implied contract list:

If the object is an item inside a list or feed, then the contract list specified by the of attribute is used.
If the object overrides (by name) an object specified in one of its contracts, then the contract list of the overridden object is used.
If all the above rules fail, then the respective primitive contract is used. For example, an obj element has an implied contract of obix:obj and real an implied contract of obj:real.

Note that element names such as bool, int, or str are syntactic sugar for an implied contract. However if an object implements one of the primitives, then it must use the correct XML element name. For example if an object implements obix:int, then it must be expressed as <int/>, rather than <obj is="obix:int"/>. Therefore it is invalid to implement multiple value types - such as implementing both obix:bool and obix:int.

6.4 Contract Inheritance

Contracts are a mechanism of inheritance – they establish the classic “is a” relationship. In the abstract sense a contract allows us to inherit a type. We can further distinguish between the explicit and implicit contract:

Explicit Contract: defines an object structure which all implementations must conform with.
Implicit Contract: defines semantics associated with the contract. Usually the implicit contract is documented using natural language prose. It isn’t mathematical, but rather subject to human interpretation.

For example when we say an object implements the Alarm contract, we immediately know that will have a child called timestamp. This structure is in the explicit contract of Alarm and is formally defined in XML. But we also attach semantics to what it means to be an Alarm object: that the object is providing information about an alarm event. These fuzzy concepts can’t be captured in machine language; rather they can only be captured in prose.

When an object declares itself to implement a contract it must meet both the explicit contract and the implicit contract. An object shouldn’t put obix:Alarm in its contract list unless it really represents an alarm event. There isn’t much more to say about implicit contracts other than it is recommended that a human brain be involved. So now let’s look at the rules governing the explicit contract.

A contract’s named children objects are automatically applied to implementations. An implementation may choose to override or default each of its contract’s children. If the implementation omits the child, then it is assumed to default to the contract’s value. If the implementation declares the child (by name), then it is overridden and the implementation’s value should be used. Let’s look at an example:

</obj>

</obj>

In this example we have a contract object identified with the URI “/def/television”. It has two children to store power and channel. Then we specify a living room TV instance that includes “/def/television” in its contract list via the is attribute. In this object, channel is overridden to 8 from its default value of 2. However since power was omitted, it is implied to default to false.

An override is always matched to its contract via the name attribute. In the example above we knew we were overriding channel, because we declared an object with a name of “channel”. We also declared an object with a name of “volume”. Since volume wasn’t declared in the contract, we assume it’s a new definition specific to this object.

Also note that the contract’s channel object declares a min and max facet. These two facets are also inherited by the implementation. Almost all attributes are inherited from their contract including facets, val, of, in, and out. The href attribute are never inherited. The null attribute inherits as follows:

If the null attribute is specified, then its explicit value is used;
If a val attribute is specified and null is unspecified, then null is implied to be false;
If neither a val attribute or a null attribute is specified, then the null attribute is inherited from the contract;

This allows us to implicitly override a null object to non-null without specifying the null attribute.

6.5 Override Rules

Contract overrides are required to obey the implicit and explicit contract. Implicit means that the implementation object provides the same semantics as the contract it implements. In the example above it would be incorrect to override channel to store picture brightness. That would break the semantic contract.

Overriding the explicit contract means to override the value, facets, or contract list. However we can never override the object to be in incompatible value type. For example if the contract specifies a child as real, then all implementations must use real for that child. As a special case, obj may be narrowed to any other element type.

We also have to be careful when overriding attributes to never break restrictions the contract has defined. Technically this means we can specialize or narrow the value space of a contract, but never generalize or widen it. This concept is called covariance. Let’s take our example from above:

In this example the contract has declared a value space of 2 to 200. Any implementation of this contract must meet this restriction. For example it would an error to override min to –100 since that would widen the value space. However we can narrow the value space by overriding min to a number greater than 2 or by overriding max to a number less than 200. The specific override rules applicable to each facet are documented in section 4.16.

6.6 Multiple Inheritance

An object’s contract list may specify multiple contract URIs to implement. This is actually quite common - even required in many cases. There are two topics associated with the implementation of multiple contracts:

Flattening: contract lists must always be flattened when specified. This comes into play when a contract has its own contract list (Section 6.6.1).
Mixins: the mixin design specifies the exact rules for how multiple contracts are merged together. This section also specifies how conflicts are handled when multiple contracts contain children with the same name (Section 6.6.2).

6.6.1 Flattening

It is common for contract objects themselves to implement contracts, just like it is common in OO languages to chain the inheritance hierarchy. However due to the nature of accessing oBIX documents over a network, we wish to minimize round trip network requests which might be required to “learn” about a complex contract hierarchy. Consider this example:

In this example if we were reading object D for the first time, it would take three more requests to fully learn what contracts are implemented (one for C, B, and A). Furthermore, if our client was just looking for objects that implemented B, it would difficult to determine this just by looking at D.

Because of these issues, servers are required to flatten their contract inheritance hierarchy into a list when specifying the is, of, in, or out attributes. In the example above, the correct representation would be:

This allows clients to quickly scan Ds contract list to see that D implements C, B, and A without further requests.

6.6.2 Mixins

Flattening is not the only reason a contract list might contain multiple contract URIs. oBIX also supports the more traditional notion of multiple inheritance using a mixin metaphor. Consider the following example:

</obj>

</obj>

</obj>

In this example ClockRadio implements both Clock and Radio. Via flattening of Clock and Radio, ClockRadio also implements Device. In oBIX this is called a mixin – Clock, Radio, and Device are mixed into (merged into) ClockRadio. Therefore ClockRadio inherits four children: serialNo, snooze, volume, and station. Mixins are a form of multiple inheritance akin to Java/C# interfaces (remember oBIX is about the type inheritance, not implementation inheritance).

Note that Clock and Radio both implement Device - the classic diamond inheritance pattern. From Device, ClockRadio inherits a child named serialNo. Furthermore notice that both Clock and Radio declare a child named volume. This naming collision could potentially create confusion for what serialNo and volume mean in ClockRadio.

In oBIX we solve this problem by flattening the contract’s children using the following rules:

Process the contract definitions in the order they are listed
If a new child is discovered, it is mixed into the object’s definition
If a child is discovered we already processed via a previous contract definition, then the previous definition takes precedence. However it is an error if the duplicate child is not contract compatible with the previous definition (see Section 6.7).

In the example above this means that Radio.volume is the definition we use for ClockRadio.volume, because Radio has a higher precedence than Clock (it is first in the contract list). Thus ClockRadio.volume has a default value of “5”. However it would be invalid if Clock.volume were declared as str, since it would not be contract compatible with Radio’s definition as an int – in that case ClockRadio could not implement both Clock and Radio. It is the server vendor’s responsibility not to create incompatible name collisions in contracts.

The first contract in a list is given specific significance since its definition trumps all others. In oBIX this contract is called the primary contract. It is recommended that the primary contract implement all the other contracts specified in the contract list (this actually happens quite naturally by itself in many programming languages). This makes it easier for clients to bind the object into a strongly typed class if desired. Obviously this recommendation doesn’t make sense for contract objects themselves – contracts shouldn’t implement themselves.

6.7 Contract Compatibility

A contract list which is covariantly substitutable with another contract list is said to be contract compatible. Contract compatibility is a useful term when talking about mixin rules and overrides for lists and operations. It is a fairly common sense notion similar to previously defined override rules – however, instead of the rules applied to individual facet attributes, we apply it to an entire contract list.

A contract list X is compatible with contract list Y, if and only if X narrows the value space defined by Y. This means that X can narrow the set of objects which implement Y, but never expand the set. Contract compatibility is not commutative (X is compatible with Y does not imply Y is compatible with X). If that definition sounds too highfaluting, you can boil it down to this practical rule: X can add new URIs to Y’s list, but never any take away.

6.8 Lists (and Feeds)

Implementations derived from list or feed contracts inherit the of attribute. Like other attributes we can override the of attribute, but only if contract compatible - you must include all of the URIs in the contract’s of attribute, but you can add additional ones (see Section 6.7).

Lists and feeds also have the special ability to implicitly define the contract list of their contents. In the following example it is implied that each child element has a contract list of /def/MissingPerson without actually specifying the is attribute in each list item:

</list>

If an element in the list or feed does specify its own is attribute, then it must be contract compatible with the of attribute.

7 XML

Chapter 4 specifies an abstract object model used to standardize how oBIX information is modeled. This chapter specifies how the object model is represented in XML.

7.1 Design Philosophy

Since there are many different approaches to developing an XML syntax, it is worthwhile to provide a bit of background to how the oBIX XML syntax was designed. Historically in M2M systems, non-standard extensions have been second class citizens at best, but usually opaque. One of the design principles of oBIX is to embrace vertical domain and vendor specific extensions, so that all data and services have a level playing field.

In order to achieve this goal, the XML syntax is designed to support a small, fixed schema for all oBIX documents. If a client agent understands this very simple syntax, then the client is guaranteed access to the server’s object tree regardless of whether those objects implement standard or non-standard contracts.

Higher level semantics are captured via contracts. Contracts “tag” an object with a type and can be applied dynamically. This is very useful for modeling systems which are dynamically configured in the field. What is important is that contracts are optionally understood by clients. Contracts do not effect the XML syntax nor are clients required to use them for basic access to the object tree. Contracts are merely an abstraction layered cleanly above the object tree and it’s fixed XML syntax.

7.2 XML Syntax

The oBIX XML syntax maps very closely to the abstract object model. The syntax is summarized:

Every oBIX object maps to exactly one XML element;
An object’s children are mapped as children XML elements;
The XML element name maps to the built-in object type;
Everything else about an object is represented as XML attributes;

The object model figure in Chapter 4 illustrates the valid XML elements and their respective attributes. Note the val object is simply an abstract base type for the objects which support the val attribute - there is no val element.

7.3 XML Encoding

The following rules apply to encoding oBIX documents:

oBIX documents must be well formed XML;
oBIX documents should begin with XML Declaration specifying their encoding;
It is strongly encouraged to use UTF-8 encoding without a byte order mark;
oBIX documents must not include a Document Type Declaration – oBIX documents cannot contain an internal or external subset;
oBIX documents should include an XML Namespace definition. Convention is declare the default namespace of the document to “http://obix.org/ns/schema/1.0”. If oBIX is embedded inside another type of XML document, then convention is to use “o” as the namespace prefix. Note that the prefix “obix” should not be used (see Section 7.6).

7.4 XML Decoding

The following rules apply to decoding of oBIX documents:

Must conform to XML processing rules as defined by XML 1.1;
Documents which are not well formed XML must be rejected;
Parsers are not required to understand a Document Type Declaration;
Any unknown element must be ignored regardless of its XML namespace
Any unknown attribute must be ignored regardless of its XML namespace

The basic rule of thumb is: strict in what you generate, and liberal in what you accept. oBIX parsers are required to ignore elements and attributes which they do not understand. However an oBIX parser should never accept an XML document which isn’t well formed (such as mismatched tags).

7.5 XML Namespace

XML namespaces for standards within the oBIX umbrella should conform to the following pattern:

http://obix.org/ns/{spec}/{version}

The XML namespace for oBIX version 1.0 is:

http://obix.org/ns/schema/1.0

All XML in this document is assumed to have this namespace unless otherwise explicitly stated.

7.6 Namespace Prefixes in Contract Lists

XML namespace prefixes defined within an oBIX document may be used to prefix the URIs of a contract list. If a URI within a contract list starts with string matching a defined XML prefix followed by the “:” colon character, then the URI is normalized by replacing the prefix with it’s namespace value.

The XML namespace prefix of “obix” is predefined. This prefix is used for all the oBIX defined contracts. The “obix” prefix is literally translated into “http://obix.org/def/”. For example the URI “obix:bool” is translated to “http://obix.org/def/bool”. Documents should not define an XML namespace using the prefix “obix” which collides with the predefined “obix” prefix – if it is defined, then it is superseded by the predefined value of “http://obix.org/def/”. All oBIX defined contracts are accessible via their HTTP URI using the HTTP binding.

An example oBIX document with XML namespace prefixes normalized:

8 Operations

Operations are the things that you can “do” to an oBIX object. They are akin to methods in traditional OO languages. Typically they map to commands rather than a variable that has continuous state. Unlike value objects which represent an object and its current state, the op element merely represents the definition of an operation you can invoke.

All operations take exactly one object as a parameter and return exactly one object as a result. The in and out attributes define the contract list for the input and output objects. If you need multiple input or output parameters, then wrap them in a single object using a contract as the signature. For example:

</obj>

Objects can override the operation definition from one of their contracts. However the new in or out contract list must be contract compatible (see Section 6.7) with the contract’s definition.

If an operation doesn’t require a parameter, then specify in as obix:Nil. If an operation doesn’t return anything, then specify out as obix:Nil. Occasionally an operation is inherited from a contract which is unsupported in the implementation. In this case use set the status attribute to disabled.

Operations are always invoked via their own href attribute (not their parent’s href). Therefore operations should always specify an href attribute if you wish clients to invoke them. A common exception to this rule is contract definitions themselves.

9 Object Composition

A good metaphor for comparison with oBIX is the World Wide Web. If you ignore all the fancy stuff like JavaScript and Flash, basically the WWW is a web of HTML documents hyperlinked together with URIs. If you dive down one more level, you could say the WWW is a web of HTML elements such as <p>, <table>, and <div>.

What the WWW does for HTML documents, oBIX does for objects. The logical model for oBIX is a global web of oBIX objects linked together via URIs. Some of these oBIX objects are static documents like contracts or device descriptions. Other oBIX objects expose real-time data or services. But they all are linked together via URIs to create the oBIX Web.

Individual objects are composed together in two ways to define this web. Objects may be composed together via containment or via reference.

9.1 Containment

Any oBIX object may contain zero or more children objects. This even includes objects which might be considered primitives such as bool or int. All objects are open ended and free to specify new objects which may not be in the object’s contract. Containment is represented in the XML syntax by nesting the XML elements:

</list>

</obj>

In this example the object identified by “/a” contains “/a/b”, which in turn contains “/a/b/c”. Child objects may be named or unnamed depending on if the name attribute is specified (Section 5.1). In the example, “/a/b” is named and “/a/b/c” is unnamed. Typically named children are used to represent fields in a record, structure, or class type. Unnamed children are often used in lists.

9.2 References

Let’s go back to our WWW metaphor. Although the WWW is a web of individual HTML elements like <p> and <div>, we don’t actually pass individual <p> elements around over the network. Rather we “chunk” them into HTML documents and always pass the entire document over the network. To tie it all together, we create links between documents using the <a> anchor element. These anchors serve as place holders, referencing outside documents via a URI.

A oBIX reference is basically just like an HTML anchor. It serves as placeholder to “link” to another oBIX object via a URI. While containment is best used to model small trees of data, references may be used to model very large trees or graphs of objects. As a matter fact, with references we can link together all oBIX objects on the Internet to create the oBIX Web.

9.3 Extents

When oBIX is applied to a problem domain, we have to decide whether to model relationships using either containment or references. These decisions have a direct impact on how your model is represented in XML and accessed over the network. The containment relationship is imbued with special semantics regarding XML encoding and eventing. In fact, oBIX coins a term for containment called an object’s extent. An object’s extent is its tree of children down to references. Only objects which have an href have an extent. Objects without an href are always included in one or more of their ancestors extents.

</obj>

</obj>

In the example above, we have five objects named ‘a’ to ‘e’. Because ‘a’ includes an href, it has an associated extent, which encompasses ‘b’ and ‘c’ by containment and ‘d’ and ‘e’ by reference. Likewise, ‘b’ has an href which results in an extent encompassing ‘c’ by containment and ‘d’ by reference. Object ‘c’ does not provide a direct href, but exists in both the ‘a’ and ‘b’ objects’ extents. Note an object with an href has exactly one extent, but can be nested inside multiple extents.

9.4 XML

When marshaling objects into an XML, it is required that an extent always be fully inlined into the XML document. The only valid objects which may be referenced outside the document are ref element themselves.

If the object implements a contract, then it is required that the extent defined by the contract be fully inlined into the document (unless the contract itself defined a child as a ref element). An example of a contract which specifies a child as a ref is Lobby.about (10.3).

10 Networking

The heart of oBIX is its object model and associated XML syntax. However, the primary use case for oBIX is to access information and services over a network. The oBIX architecture is based on a client/server network model:

Server: software containing oBIX enabled data and services. Servers respond to requests from client over a network.
Client: software which makes requests to servers over a network to access oBIX enabled data and services.

There is nothing to prevent software from being both an oBIX client and server. Although a key tenant of oBIX is that a client is not required to implement server functionality which might require a server socket to accept incoming requests.

10.1 Request / Response

All network access is boiled down into three request / response types:

Read: return the current state of an object at a given URI as an oBIX XML document.
Write: update the state of an existing object at a URI. The state to write is passed over the network as an oBIX XML document. The new updated state is returned in an oBIX XML document.
Invoke: invoke an operation identified by a given URI. The input parameter and output result are passed over the network as an oBIX XML document.

Exactly how these three request/responses are implemented between a client and server is called a protocol binding. The oBIX specification defines two standard protocol bindings: HTTP Binding (see Chapter 17) and SOAP Binding (see Chapter 18). However all protocol bindings must follow the same read, write, invoke semantics discussed next.

10.1.1 Read

The read request specifies an object’s URI and the read response returns the current state of the object as an oBIX document. The response must include the object’s complete extent (see 9.3). Servers may return an err object to indicate the read was unsuccessful – the most common error is obix:BadUriErr (see 10.2 for standard error contracts).

10.1.2 Write

The write request is designed to overwrite the current state of an existing object. The write request specifies the URI of an existing object and it’s new desired state. The response returns the updated state of the object. If the write is successful, the response must include the object’s complete extent (see 9.3). If the write is unsuccessful, then the server must return an err object indicating the failure.

The server is free to completely or partially ignore the write, so clients should be prepared to examine the response to check if the write was successful. Servers may also return an err object to indicate the write was unsuccessful.

Clients are not required to include the object’s full extent in the request. Objects explicitly specified in the request object tree should be overwritten or “overlaid” over the server’s actual object tree. Only the val attribute should be specified for a write request (outside of identification attributes such as name). A write operation that provides facets has unspecified behavior. When writing int or reals with units, the write value must be in the same units as the server specifies in read requests – clients must not provide a different unit facet and expect the server to auto-convert (in fact the unit facet should be not included in the request).

10.1.3 Invoke

The invoke request is designed to trigger an operation. The invoke request specified the URI of an op object and the input argument object. The response includes the output object. The response must include the output object’s complete extent (see 9.3). Servers may also return an err object to indicate the invoke was unsuccessful.

10.2 Errors

Request errors are conveyed to clients with the err element. Any time an oBIX server successfully receives a request and the request cannot be processed, then the server should return an err object to the client. Returning a valid oBIX document with err must be used when feasible rather than protocol specific error handling (such as an HTTP response code). Such a design allows for consistency with batch request partial failures and makes protocol binding more pluggable by separating data transport from application level error handling.

A few contracts are predefined for common errors:

BadUriErr: used to indicate either a malformed URI or a unknown URI;
UnsupportedErr: used to indicate an a request which isn’t supported by the server implementation (such as an operation defined in a contract, which the server doesn’t support);
PermissionErr: used to indicate that the client lacks the necessary security permission to access the object or operation.

The contracts for these errors are:

If one of the above contracts makes sense for an error, then it should be included in the err element’s is attribute. It is strongly encouraged to also include a useful description of the problem in the display attribute.

10.3 Lobby

All oBIX servers must provide an object which implements obix:Lobby. The Lobby object serves as the central entry point into an oBIX server, and lists the URIs for other well-known objects defined by the oBIX specification. Theoretically all a client needs to know to bootstrap discovery is one URI for the Lobby instance. By convention this URI is “http://server/obix”, although vendors are certainly free to pick another URI. The Lobby contract is:

</obj>

The Lobby instance is where vendors should place vendor specific objects used for data and service discovery.

10.4 About

The obix:About object is a standardized list of summary information about an oBIX server. Clients can discover the About URI directly from the Lobby. The About contract is:

</obj>

The following children provide information about the oBIX implementation:

obixVersion: specifies which version of the oBIX specification the server implements. This string must be a list of decimal numbers separated by the dot character (Unicode 0x2E).

The following children provide information about the server itself:

serverName: provides a short localized name for the server.
serverTime: provides the server’s current local time.
serverBootTime: provides the server’s start time - this should be the start time of the oBIX server software, not the machine’s boot time.

The following children provide information about the server’s software vendor:

vendorName: the company name of the vendor who implemented the oBIX server software.
vendorUrl: a URI to the vendor’s website.

The following children provide information about the software product running the server:

productName: with the product name of oBIX server software.
productUrl: a URI to the product’s website.
productVersion: a string with the product’s version number. Convention is to use decimal digits separated by dots.

10.5 Batch

The Lobby defines a batch operation which is used to batch multiple network requests together into a single operation. Batching multiple requests together can often provide significant performance improvements over individual round-robin network requests. As a general rule, one big request will always out-perform many small requests over a network.

A batch request is an aggregation of read, write, and invoke requests implemented as a standard oBIX operation. At the protocol binding layer, it is represented as a single invoke request using the Lobby.batch URI. Batching a set of requests to a server must be processed semantically equivalent to invoking each of the requests individually in a linear sequence.

The batch operation inputs a BatchIn object and outputs a BatchOut object:

The BatchIn contract specifies a list of requests to process identified using the Read, Write, or Invoke contract:

</uri>

</uri>

The BatchOut contract specifies an ordered list of the response objects to each respective request. For example the first object in BatchOut must be the result of the first request in BatchIn. Failures are represented using the err object. Every uri passed via BatchIn for a read or write request must have a corresponding result obj in BatchOut with an href attribute using an identical string representation from BatchIn (no normalization or case conversion is allowed).

It is up to vendors to decide how to deal with partial failures. In general idempotent requests should indicate a partial failure using err, and continue processing additional requests in the batch. If a server decides not to process additional requests when an error is encountered, then it is still required to return an err for each respective request not processed.

Let’s look at a simple example:

</uri>

</list>

</list>

In this example, the batch request is specifying a read request for “/someStr” and “/invalidUri”, followed by a write request to “/someStr”. Note that the write request includes the value to write as a child named “in”.

The server responds to the batch request by specifying exactly one object for each request URI. The first read request returns a str object indicating the current value identified by “/someStr”. The second read request contains an invalid URI, so the server returns an err object indicating a partial failure and continues to process subsequent requests. The third request is a write to “someStr”. The server updates the value at “someStr”, and returns the new value. Note that because the requests are processed in order, the first request provides the original value of “someStr” and the third request contains the new value. This is exactly what we would expect had we processed each of these requests individually.

11 Core Contract Library

This chapter defines some fundamental object contracts that serve as building blocks for the oBIX specification.

11.1 Nil

The obix:Nil contract defines a standardized null object. Nil is commonly used for an operation’s in or out attribute to denote the absence of an input or output. The definition:

11.2 Range

The obix:Range contract is used to define an bool or enum’s range. Range is a list object that contains zero or more objects called the range items. Each item’s name attribute specifies the identifier used as the literal value of an enum. Item ids are never localized, and must be used only once in a given range. You may use the optional displayName attribute to specify a localized string to use in a user interface. The definition of Range:

An example:

</list>

The range facet may be used to define the localized text of a bool value using the ids of “true” and “false”:

</list >

11.3 Weekday

The obix:Weekday contract is a standardized enum for the days of the week:

</list>

</enum>

11.4 Month

The obix:Month contract is a standardized enum for the months of the year:

</list>

</enum>

11.5 Units

Representing units of measurement in software is a thorny issue. oBIX provides a unit framework for mathematically defining units within the object model. An extensive database of predefined units is also provided.

All units measure a specific quantity or dimension in the physical world. Most known dimensions can be expressed as a ratio of the seven fundamental dimensions: length, mass, time, temperature, electrical current, amount of substance, and luminous intensity. These seven dimensions are represented in SI respectively as kilogram (kg), meter (m), second (sec), Kelvin (K), ampere (A), mole (mol), and candela (cd).

The obix:Dimension contract defines the ratio of the seven SI units using a positive or negative exponent:

</obj>

A Dimension object contains zero or more ratios of kg, m, sec, K, A, mol, or cd. Each of these ratio maps to the exponent of that base SI unit. If a ratio is missing then the default value of zero is implied. For example acceleration is m/s², which would be encoded in oBIX as:

</obj>

Units with equal dimensions are considered to measure the same physical quantity. This is not always precisely true, but is good enough for practice. This means that units with the same dimension are convertible. Conversion can be expressed by specifying the formula required to convert the unit to the dimension’s normalized unit. The normalized unit for every dimension is the ratio of SI units itself. For example the normalized unit of energy is the joule m²·kg·s^-2. The kilojoule is 1000 joules and the watt-hour is 3600 joules. Most units can be mathematically converted to their normalized unit and to other units using the linear equations:

unit = dimension · scale + offset

toNormal = scalar · scale + offset

fromNormal = (scalar - offset) / scale

toUnit = fromUnit.fromNormal( toUnit.toNormal(scalar) )

There are some units which don’t fit this model including logarithm units and units dealing with angles. But this model provides a practical solution for most problem spaces. Units which don’t fit this model should use a dimension where every exponent is set to zero. Applications should not attempt conversions on these types of units.

The obix:Unit contract defines a unit including its dimension and its toNormal equation:

</obj>

The unit element contains a symbol, dimension, scale, and offset sub-object:

symbol: The symbol element defines a short abbreviation to use for the unit. For example “°F” would be the symbol for degrees Fahrenheit. The symbol element should always be specified.
dimension: The dimension object defines the dimension of measurement as a ratio of the seven base SI units. If omitted, the dimension object defaults to the obix:Dimension contract, in which case the ratio is the zero exponent for all seven base units.
scale: The scale element defines the scale variable of the toNormal equation. The scale object defaults to 1.
offset: The offset element defines the offset variable of the toNormal equation. If omitted then offset defaults to 0.

The display attribute should be used to provide a localized full name for the unit based on the client’s locale. If the display attribute is omitted, clients should use symbol for display purposes.

An example for the predefined unit for kilowatt:

</obj>

</obj>

Automatic conversion of units is considered a localization issue – see Section 17.3 for more details.

12 Watches

A key requirement of oBIX is access to real-time information. We wish to enable clients to efficiently receive access to rapidly changing data. However, we don’t want to require clients to implement web servers or expose a well-known IP address. In order to address this problem, oBIX provides a model for client polled eventing called watches. The watch lifecycle is as follows:

The client creates a new watch object with the make operation on the server’s WatchService URI. The server defines a new Watch object and provides a URI to access the new watch.
The client registers (and unregisters) objects to watch using operations on the Watch object.
The client periodically polls the Watch URI using the pollChanges operation to obtain the events which have occurred since the last poll.
The server frees the Watch under two conditions. The client may explicitly free the Watch using the delete operation. Or the server may automatically free the Watch because the client fails to poll after a predetermined amount of time (called the lease time).

Watches allow a client to maintain a real-time cache for the current state of one or more objects. They are also used to access an event stream from a feed object. Plus, watches serve as the standardized mechanism for managing per-client state on the server via leases.

12.1 WatchService

The WatchService object provides a well-known URI as the factory for creating new watches. The WatchService URI is available directly from the Lobby object. The contract for WatchService:

</obj>

The make operation returns a new empty Watch object as an output. The href of the newly created Watch object can then be used for invoking operations to populate and poll the data set.

12.2 Watch

Watch object is used to manage a set of objects which are subscribed and periodically polled by clients to receive the latest events. The contract is:

</obj>

</obj>

</uri>

</obj>

Many of the Watch operations use two contracts: obix:WatchIn and obix:WatchOut. The client identifies objects to add and remove from the poll list via WatchIn. This object contains a list of URIs. Typically these URIs should be server relative.

The server responds to add, pollChanges, and pollRefresh operations via the WatchOut contract. This object contains the list of subscribed objects - each object must specify an href URI using the exact same string as the URI identified by the client in the corresponding WatchIn. Servers are not allowed to perform any case conversions or normalization on the URI passed by the client. This allows client software to use the URI string as a hash key to match up server responses.

12.2.1 Watch.add

Once a Watch has been created, the client can add new objects to watch using the add operation. This operation inputs a list of URIs and outputs the current value of the objects referenced. The objects returned are required to specify an href using the exact string representation input by the client. If any object cannot be processed, then a partial failure should be expressed by returning an err object with the respective href. Subsequent URIs must not be effected by the failure of one invalid URI. The add operation should never return objects not explicitly included in the input URIs (even if there are already existing objects in the watch list). No guarantee is made that the order of objects in WatchOut match the order in of URIs in WatchIn – clients must use the URI as a key for matching.

Note that the URIs supplied via WatchIn may include an optional in parameter. This parameter is only used when subscribing a watch to a feed object. Feeds also differ from other objects in that they return a list of historic events in WatchOut. Feeds are discussed in detail in Section 12.4.

It is invalid to add an op’s href to a watch, the server must report an err.

If an attempt is made to add a URI to a watch which was previously already added, then the server should return the current object’s value in the WatchOut result, but treat poll operations as if the URI was only added once – polls should only return the object once. If an attempt is made to add the same URI multiple times in the same WatchIn request, then the server should only return the object once.

Note: the lack of a trailing slash can cause problems with watches. Consider a client which adds a URI to a watch without a trailing slash. The client will use this URI as a key in its local hashtable for the watch. Therefore the server must use the URI exactly as the client specified. However, if the object’s extent includes children objects they will not be able to use relative URIs. It is recommended that servers fail-fast in these cases and return a BadUriErr when clients attempt to add a URI without a trailing slash to a watch (even though they may allow it for a normal read request).

12.2.2 Watch.remove

The client can remove objects from the watch list using the remove operation. A list of URIs is input to remove, and the Nil object is returned. Subsequent pollChanges and pollRefresh operations must cease to include the specified URIs. It is possible to remove every URI in the watch list; but this scenario must not automatically free the Watch, rather normal poll and lease rules still apply. It is invalid to use the WatchInItem.in parameter for a remove operation.

12.2.3 Watch.pollChanges

Clients should periodically poll the server using the pollChanges operation. This operation returns a list of the subscribed objects which have changed. Servers should only return the objects which have been modified since the last poll request for the specific Watch. As with add, every object must specify an href using the exact same string representation the client passed in the original add operation. The entire extent of the object should be returned to the client if any one thing inside the extent has changed on the server side.

Invalid URIs must never be included in the response (only in add and pollRefresh). An exception to this rule is when an object which is valid is removed from the URI space. Servers should indicate an object has been removed via an err with the BadUriErr contract.

12.2.4 Watch.pollRefresh

The pollRefresh operation forces an update of every object in the watch list. The server must return every object and it’s full extent in the response using the href with the exact same string representation passed by the client in the original add. Invalid URIs in the poll list should be included in the response as an err element. A pollRefresh resets the poll state of every object, so that the next pollChanges only returns objects which have changed state since the pollRefresh invocation.

12.2.5 Watch.lease

All Watches have a lease time, specified by the lease child. If the lease time elapses without the client initiating a request on the Watch, then the server is free to expire the watch. Every new poll request resets the lease timer. So as long as the client polls at least as often as the lease time, the server should maintain the Watch. The following requests should reset the lease timer: read of the Watch URI itself or invocation of the add, remove, pollChanges, or pollRefresh operations.

Clients may request a difference lease time by writing to the lease object (requires servers to assign an href to the lease child). The server is free to honor the request, cap the lease within a specific range, or ignore the request. In all cases the write request will return a response containing the new lease time in effect.

Servers should report expired watches by returning an err object with the BadUriErr contract. As a general principle servers should honor watches until the lease runs out or the client explicitly invokes delete. However, servers are free to cancel watches as needed (such as power failure) and the burden is on clients to re-establish a new watch.

12.2.6 Watch.delete

The delete operation can be used to cancel an existing watch. Clients should always delete their watch when possible to be good oBIX citizens. However servers should always cleanup correctly without an explicit delete when the lease expires.

12.3 Watch Depth

When a watch is put on an object which itself has children objects, how does a client know how “deep” the subscription goes? oBIX requires watch depth to match an object‘s extent (see Section 9.3). When a watch is put on a target object, a server must notify the client of any changes to any of the objects within that target object’s extent. If the extent includes feed objects they are not included in the watch – feeds have special watch semantics discussed in Section 12.4. This means a watch is inclusive of all descendents within the extent except refs and feeds.

12.4 Feeds

Servers may expose event streams using the feed object. The event instances are typed via the feed’s of attribute. Clients subscribe to events by adding the feed’s href to a watch, optionally passing an input parameter which is typed via the feed’s in attribute. The object returned from Watch.add is a list of historic events (or the empty list if no event history is available). Subsequent calls to pollChanges returns the list of events which have occurred since the last poll.

Let’s consider a simple example for an object which fires an event when its geographic location changes:

<obj>

</obj>

We subscribe to the moved event feed by adding “/car/moved” to a watch. The WatchOut will include the list of any historic events which have occurred up to this point in time. If the server does not maintain an event history this list will be empty:

</list>

</obj>

</list>

</obj>

Now every time we call pollChanges for the watch, the server will send us the list of event instances which have accumulated since our last poll:

<obj>

</obj>

<obj>

</obj>

</feed>

</list>

</obj>

Note the feed’s of attribute works just like the list’s of attribute. The children event instances are assumed to inherit the contract defined by of unless explicitly overridden. If an event instance does override the of contract, then it must be contract compatible. Refer to the rules defined in Section 6.8.

Invoking a pollRefresh operation on a watch with a feed that has an event history, should return all the historical events as if the pollRefresh was an add operation. If an event history is not available, then pollRefresh should act like a normal pollChanges and just return the events which have occurred since the last poll.

13 Points

Anyone familiar with automation systems immediately identifies with the term point (sometimes called tags in the industrial space). Although there are many different definitions, generally points map directly to a sensor or actuator (called hard points). Sometimes the concept of a point is mapped to a configuration variable such as a software setpoint (called soft points). In some systems point is an atomic value, and in others it encapsulates a whole truckload of status and configuration information.

The goal of oBIX is to capture a normalization representation of points without forcing an impedance mismatch on vendors trying to make their native system oBIX accessible. To meet this requirement, oBIX defines a low level abstraction for point - simply one of the primitive value types with associated status information. Point is basically just a marker contract used to tag an object as exhibiting “point” semantics:

This contract must only be used with the value primitive types: bool, real, enum, str, abstime, and reltime. Points should use the status attribute to convey quality information. The following table specifies how to map common control system semantics to a value type:

bool	digital point	<bool is="obix:Point" val="true"/>
real	analog point	<real is="obix:Point" val="22" units="obix:units/celsius"/>
enum	multi-state point	<enum is="obix:Point" val="slow"/>

13.1 Writable Points

Different control systems handle point writes using a wide variety of semantics. Sometimes we write a point at a specific priority level. Sometimes we override a point for a limited period of time, after which the point falls back to a default value. The oBIX specification doesn’t attempt to impose a specific model on vendors. Rather oBIX provides a standard WritablePoint contract which may be extended with additional mixins to handle special cases. WritablePoint defines write as an operation which takes a WritePointIn structure containing the value to write. The contracts are:

</obj>

</obj>

It is implied that the value passed to writePoint match the type of the point. For example if WritablePoint is used with an enum, then writePoint must pass an enum for the value.

14 History

Most automation systems have the ability to persist periodic samples of point data to create a historical archive of a point’s value over time. This feature goes by many names including logs, trends, or histories. In oBIX, a history is defined as a list of time stamped point values. The following features are provided by oBIX histories:

History Object: a normalized representation for a history itself;
History Record: a record of a point sampling at a specific timestamp
History Query: a standard way to query history data as Points;
History Rollup: a standard mechanism to do basic rollups of history data;

14.1 History Object

Any object which wishes to expose itself as a standard oBIX history implements the obix:History contract:

</obj>

Let’s look at each of History’s sub-objects:

count: this field stores the number of history records contained by the history;
start: this field provides the timestamp of the oldest record;
end: this field provides the timestamp of the newest record;
query: the query object is used to query the history to read history records;
feed: used to subscribe to a real-time feed of history records;
rollup: this object is used to perform history rollups (it is only supported for numeric history data);

An example of a history which contains an hour of 15 minute temperature data:

</obj>

14.2 History Queries

Every History object contains a query operation to query the historical data.

14.2.1 HistoryFilter

The History.query input contract:

</obj>

These fields are described in detail:

limit: an integer indicating the maximum number of records to return. Clients can use this field to throttle the amount of data returned by making it non-null. Servers must never return more records than the specified limit. However servers are free to return fewer records than the limit.
start: if non-null this field indicates an inclusive lower bound for the query’s time range.
end: if non-null this field indicates an inclusive upper bound for the query’s time range.

14.2.2 HistoryQueryOut

The History.query output contract:

</obj>

Just like History, every HistoryQueryOut returns count, start, and end. But unlike History, these values are for the query result, not the entire history. The actual history data is stored as a list of HistoryRecords in the data field. Remember that child order is not guaranteed in oBIX, therefore it might be common to have count after data.

14.2.3 HistoryRecord

The HistoryRecord contract specifies a record in a history query result:

</obj>

Typically the value should be on the value types used with obix:Point.

14.2.4 History Query Example

An example query from the “/outsideAirTemp/history” example above:

</list>

</obj>

Note in the example above how the data list uses a document local contract to define facets common to all the records (although we still have to flatten the contract list).

14.3 History Rollups

Control systems collect historical data as raw time sampled values. However, most applications wish to consume historical data in a summarized form which we call rollups. The rollup operation is used summarize an interval of time. History rollups only apply to histories which store numeric information as a list of RealPoints. Attempting to query a rollup on a non-numeric history such as a history of BoolPoints should result in an error.

14.3.1 HistoryRollupIn

The History.rollup input contract extends HistoryFilter to add an interval parameter:

</obj>

14.3.2 HistoryRollupOut

The History.rollup output contract:

</obj>

The HistoryRollupOut object looks very much like HistoryQueryOut except it returns a list of HistoryRollupRecords, rather than HistoryRecords. Note: unlike HistoryQueryOut, the start for HistoryRollupOut is exclusive, not inclusive. This issue is discussed in greater detail next.

14.3.3 HistoryRollupRecord

A history rollup returns a list of HistoryRollupRecords:

</obj>

The children are defined as:

start: the exclusive start time of the record’s rollup interval;
end: the inclusive end time of the record’s rollup interval;
count: the number of records used to compute this rollup interval;
min: specifies the minimum value of all the records within the interval;
max: specifies the maximum value of all the records within the interval;
avg: specifies the mathematical average of all the values within the interval;
sum: specifies the summation of all the values within the interval;

14.3.4 Rollup Calculation

The best way to understand how rollup calculations work is through an example. Let’s consider a history of meter data where we collected two hours of 15 minute readings of kilowatt values:

</list>

<obj>

</obj>

If we were to query the rollup using an interval of 1 hour with a start time of 12:00 and end time of 14:00, the result should be:

<obj>

</obj>

<obj>

</obj>

</list>

</obj>

If you whip out your calculator, the first thing you will note is that the first raw record of 80kW was never used in the rollup. This is because start time is always exclusive. The reason start time has to be exclusive is because we are summarizing discrete samples into a contiguous time range. It would be incorrect to include a record in two different rollup intervals! To avoid this problem we always make start time exclusive and end time inclusive. The following table illustrates how the raw records were applied to rollup intervals:

Interval Start (exclusive)	Interval End (inclusive)	Records Included
2005-03-16T12:00	2005-03-16T13:00	82 + 90 + 85 + 81 = 338
2005-03-16T13:00	2005-03-16T14:00	84 + 91 + 83 + 78 = 336

14.4 History Feeds

The History contract specifies a feed for subscribing to a real-time feed of the history records. History.feed reuses the same HistoryFilter input contract used by History.query – the same semantics apply. When adding a History feed to a watch, the initial result should contain the list of HistoryRecords filtered by the input parameter (the initial result should match what History.query would return). Subsequent calls to Watch.pollChanges should return any new HistoryRecords which have been collected since the last poll that also satisfy the HistoryFilter.

15 Alarming

The oBIX alarming feature specifies a normalized model to query, watch, and acknowledge alarms. In oBIX, an alarm indicates a condition which requires notification of either a user or another application. In many cases an alarm requires acknowledgement, indicating that someone (or something) has taken action to resolve the alarm condition. The typical lifecycle of an alarm is:

Source Monitoring: algorithms in a server monitor an alarm source. An alarm source is an object with an href which has the potential to generate an alarm. Example of alarm sources might include sensor points (this room is too hot), hardware problems (disk is full), or applications (building is consuming too much energy at current energy rates)
Alarm Generation: if the algorithms in the server detect that an alarm source has entered an alarm condition, then an alarm record is generated. Every alarm is uniquely identified using an href and represented using the obix:Alarm contract. Sometimes we refer to the alarm transition as off-normal.
To Normal: many alarm sources are said to be stateful - eventually the alarm source exits the alarm state, and is said to return to-normal. Stateful alarms implement the obix:StatefulAlarm contract. When the source transitions to normal, we update normalTimestamp of the alarm.
Acknowledgement: often we require that a user or application acknowledges that they have processed an alarm. These alarms implement the obix:AckAlarm contract. When the alarm is acknowledged, we update ackTimestamp and ackUser.

15.1 Alarm States

Alarm state is summarized with two variables:

In Alarm: is the alarm source currently in the alarm condition or in the normal condition. This variables maps to the alarm status state.
Acknowledged: is the alarm acknowledged or unacknowledged. This variable maps to the unacked status state.

Either of these states may transition independent of the other. For example an alarm source can return to normal before or after an alarm has been acknowledged. Furthermore it is not uncommon to transition between normal and off-normal multiple times generating several alarm records before any acknowledgements occur.

Note not all alarms have state. An alarm which implements neither StatefulAlarm nor the AckAlarm contracts is completely stateless – these alarms merely represent event. An alarm which implements StatefulAlarm but not AckAlarm will have an in-alarm state, but not acknowledgement state. Conversely an alarm which implements AckAlarm but not StatefulAlarm will have an acknowledgement state, but not in-alarm state.

15.1.1 Alarm Source

The current alarm state of an alarm source is represented using the status attribute. This attribute is discussed in Section 4.16.8. It is recommended that alarm sources always report their status via the status attribute.

15.1.2 StatefulAlarm and AckAlarm

An Alarm record is used to summarize the entire lifecycle of an alarm event. If the alarm implements StatefulAlarm it tracks transition from off-normal back to normal. If the alarm implements AckAlarm, then it also summarizes the acknowledgement. This allows for four discrete alarm states:

alarm	acked	normalTimestamp	ackTimestamp
true	false	null	null
true	true	null	non-null
false	false	non-null	null
false	true	non-null	non-null

15.2 Alarm Contracts

15.2.1 Alarm

The core Alarm contract is:

</obj>

The child objects are:

source: the URI which identifies the alarm source. The source should reference an oBIX object which models the entity that generated the alarm.
timestamp: this is the time at which the alarm source transitioned from normal to off-normal and the Alarm record was created.

15.2.2 StatefulAlarm

Alarms which represent an alarm state which may transition back to normal must implement the StatefulAlarm contract:

</obj>

The child object is:

normalTimestamp: if the alarm source is still in the alarm condition, then this field is null. Otherwise this indicates the time of the transition back to the normal condition.

15.2.3 AckAlarm

Alarms which support acknowledgement must implement the AckAlarm contract:

</obj>

</obj>

</obj>

The child objects are:

ackTimestamp: if the alarm is unacknowledged, then this field is null. Otherwise this indicates the time of the acknowledgement.
ackUser: if the alarm is unacknowledged, then this field is null. Otherwise this field should provide a string indicating who was responsible for the acknowledgement.

The ack operation is used to programmatically acknowledge the alarm. The client may optionally specify an ackUser string via AlarmAckIn. However, the server is free to ignore this field depending on security conditions. For example a highly trusted client may be allowed to specify its own ackUser, but a less trustworthy client may have its ackUser predefined based on the authentication credentials of the protocol binding. The ack operation returns an AckAlarmOut which contains the updated alarm record. Use the Lobby.batch operation to efficiently acknowledge a set of alarms.

15.2.4 PointAlarms

It is very common for an alarm source to be an obix:Point. A respective PointAlarm contract is provided as a normalized way to report the value which caused the alarm condition:

</obj>

The alarmValue object should be one of the value types defined for obix:Point in Section 13.

15.3 AlarmSubject

Servers which implement oBIX alarming must provide one or more objects which implement the AlarmSubject contract. The AlarmSubject contract provides the ability to categorize and group the sets of alarms a client may discover, query, and watch. For instance a server could provide one AlarmSubject for all alarms and other AlarmSubjects based on priority or time of day. The contract for AlarmSubject is:

</obj>

</obj>

</obj>

The AlarmSubject follows the same design pattern as History. The AlarmSubject specifies the active count of alarms; however, unlike History it does not provide the start and end bounding timestamps. It contains a query operation to read the current list of alarms with an AlarmFilter to filter by time bounds. AlarmSubject also contains a feed object which may be used to subscribe to the alarm events.

15.4 Alarm Feed Example

The following example illustrates how a feed works with this AlarmSubject:

</obj>

The server indicates it has two open alarms under the specified AlarmSubject. If a client were to add the AlarmSubject’s feed to a watch:

</list>

</obj>

</obj>

</obj>

</feed>

</list>

</obj>

The watch returns the historic list of alarm events which is two open alarms. The first alarm indicates an out of bounds condition in AirHandler-2’s return temperature. The second alarm indicates that the system has detected that the front door has been propped open.

Now let’s fictionalize that the system detects the front door is closed, and alarm point transitions to the normal state. The next time the client polls the watch the alarm would show up in the feed list (along with any additional changes or new alarms not shown here):

</obj>

</feed>

</list>

</obj>

16 Security

Security is a broad topic, that covers many issues:

Authentication: verifying a user (client) is who he says he is;
Encryption: protecting oBIX documents from prying eyes;
Permissions: checking a user’s permissions before granting access to read/write objects or invoke operations;
User Management: managing user accounts and permissions levels;

The basic philosophy of oBIX is to leave these issues outside of the specification. Authentication and encryption is left as a protocol binding issue. Privileges and user management is left as a vendor implementation issue. Although it is entirely possible to define a publicly exposed user management model through oBIX, this specification does not define any standard contracts for user management.

16.1 Error Handling

It is expected that an oBIX server will perform authentication and utilize those user credentials for checking permissions before processing read, write, and invoke requests. As a general rule, servers should return err with the obix:PermissionErr contract to indicate a client lacks the permission to perform a request. In particularly sensitive applications, a server may instead choose to return BadUriErr so that an untrustworthy client is unaware that a specific object even exists.

16.2 Permission based Degradation

Servers should strive to present their object model to a client based on the privileges available to the client. This behavior is called permission based degradation. The following rules summarize effective permission based degradation:

If an object cannot be read, then it should not be discoverable through objects which are available.
Servers should attempt to group standard contracts within the same privilege level – for example don’t split obix:History’s start and end into two different security levels such that a client might be able to read start, and not end.
Don’t include a contract in an object’s is attribute if the contract’s children are not readable to the client.
If an object isn’t writable, then make sure the writable attribute is set to false (either explicitly or through a contract default).
If an op inherited from a visible contract cannot be invoked, then set the null attribute to true to disable it.

17 HTTP Binding

The HTTP binding specifies a simple REST mapping of oBIX requests to HTTP. A read request is a simple HTTP GET, which means that you can simply read an object by typing its URI into your browser. Refer to “RFC 2616 Hypertext Transfer Protocol” for the full specification of HTTP 1.1.

17.1 Requests

The following table summarizes how oBIX requests map to HTTP methods:

oBIX Request	HTTP Method	Target
Read	GET	Any object with an href
Write	PUT	Any object with an href and writable=true
Invoke	POST	Any op object

The URI used for an HTTP request must map to the URI of the object being read, written, or invoked. Read requests use a simple HTTP GET and return the resulting oBIX document. Write and invoke are implemented with the PUT and POST methods respectively. The input is passed to the server as an oBIX document and the result is returned as an oBIX document.

If the oBIX server processes a request, then it must return the resulting oBIX document with an HTTP status code of 200 OK. The 200 status code must be used even if the request failed and the server is returning an err object as the result.

The oBIX documents passed between client and servers should specify a MIME type of “text/xml” for the Content-Type HTTP header.

Clients and servers must encode the oBIX document passed over the network using standard XML encoding rules. It is strongly recommended using UTF8 without a byte-order mark. If specified, the Content-Encoding HTTP header must match the XML encoding.

17.2 Security

Numerous standards are designed to provide authentication and encryption services for HTTP. Existing standards should be used when applicable for oBIX HTTP implementations including:

RFC 2617 - HTTP Authentication: Basic and Digest Access Authentication
RFC 2818 - HTTP Over TLS (HTTPS)
RFC 4346/2246 – The TLS Protocol (Transport Layer Security)

17.3 Localization

Servers should localize appropriate data based on the desired locale of the client agent. Localization should include the display and displayName attributes. The desired locale of the client should be determined through authentication or via the Accept-Language HTTP header. A suggested algorithm is to check if the authenticated user has a preferred locale configured in the server’s user database, and if not then fallback to the locale derived from the Accept-Language header.

Localization may include auto-conversion of units. For example if the authenticated user has a configured a preferred unit system such as English versus Metric, then the server might attempt to convert values with an associated unit facet to the desired unit system.

18 SOAP Binding

The SOAP binding maps a SOAP operation to each of the three oBIX request types: read, write and invoke. Like the HTTP binding, read is supported by every object, write is supported by objects whose writable attribute is true, and invoke is only supported by operations. Inputs and outputs of each request are specific to the target object.

Unlike the HTTP binding, requests are not accessed via the URI of the target object, but instead via the URI of the SOAP server with the object’s URI encoded into the body of the SOAP envelope.

18.1 SOAP Example

The following is a SOAP request to an oBIX server’s About object:

<env:Envelope xmlns:env="http://schemas.xmlsoap.org/soap/envelope/">

<env:Body>

<read xmlns="http://obix.org/ns/wsdl/1.0"

href="http://localhost/obix/about" />

</env:Body>

</env:Envelope>

An example response to the above request:

<env:Envelope xmlns:env="http://schemas.xmlsoap.org/soap/envelope/">

<env:Body>

<obj name="about"

href="http://localhost/obix/about/"

xmlns="http://obix.org/ns/schema/1.0">

</obj>

</env:Body>

</env:Envelope>

18.2 Error Handling

The oBIX specification defines no SOAP faults. If a request is processed by an oBIX server, then a valid oBIX document should be returned with a failure indicated via the err object.

18.3 Security

Refer to the recommendations in WS-I Basic Profile 1.0 for security:

http://www.ws-i.org/Profiles/BasicProfile-1.0-2004-04-16.html#security

18.4 Localization

SOAP bindings should follow localization patterns defined for the HTTP binding when applicable (see Section 17.3).

18.5 WSDL

In the types section of the WSDL document, the oBIX schema is imported. Server implementations might consider providing the schemaLocation attribute which is absent in the standard document.

Missing from the standard oBIX WSDL is the service element. This element binds a SOAP server instance with a network address. Each instance will have to provide its own services section of the WSDL document. The following is an example of the WSDL service element:

<wsdl:service name="obix">

<wsdl:port name="obixPort" binding="tns:obixSoapBinding">

<soap:address location="http://localhost/obix/soap"/>

</wsdl:port>

</wsdl:service>

Standard oBIX WSDL is:

<wsdl:definitions targetNamespace="http://obix.org/ns/wsdl/1.0"

xmlns="http://obix.org/ns/wsdl/1.0"

xmlns:wsdl="http://schemas.xmlsoap.org/wsdl/"

xmlns:soap="http://schemas.xmlsoap.org/wsdl/soap/"

xmlns:xsd="http://www.w3.org/2001/XMLSchema"

xmlns:obix="http://obix.org/ns/schema/1.0">

<wsdl:types>

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema"

elementFormDefault="qualified"

targetNamespace="http://obix.org/ns/wsdl/1.0">

<xsd:import namespace="http://obix.org/ns/schema/1.0"/>

<xsd:complexType name="ReadReq">

<xsd:attribute name="href" type="xsd:anyURI"/>

</xsd:complexType>

<xsd:complexType name="WriteReq">

<xsd:complexContent>

<xsd:extension base="ReadReq">

<xsd:sequence>

<xsd:element ref="obix:obj" maxOccurs="1" minOccurs="1"/>

</xsd:sequence>

</xsd:extension>

</xsd:complexContent>

</xsd:complexType>

<xsd:complexType name="InvokeReq">

<xsd:complexContent>

<xsd:extension base="ReadReq">

<xsd:sequence>

<xsd:element ref="obix:obj" maxOccurs="1" minOccurs="1"/>

</xsd:sequence>

</xsd:extension>

</xsd:complexContent>

</xsd:complexType>

<xsd:element name="read" type="ReadReq"/>

<xsd:element name="write" type="WriteReq"/>

<xsd:element name="invoke" type="InvokeReq"/>

</xsd:schema>

</wsdl:types>

<wsdl:message name="readSoapReq">

<wsdl:part name="body" element="read"/>

</wsdl:message>

<wsdl:message name="readSoapRes">

<wsdl:part name="body" element="obix:obj"/>

</wsdl:message>

<wsdl:message name="writeSoapReq">

<wsdl:part name="body" element="write"/>

</wsdl:message>

<wsdl:message name="writeSoapRes">

<wsdl:part name="body" element="obix:obj"/>

</wsdl:message>

<wsdl:message name="invokeSoapReq">

<wsdl:part name="body" element="invoke"/>

</wsdl:message>

<wsdl:message name="invokeSoapRes">

<wsdl:part name="body" element="obix:obj"/>

</wsdl:message>

<wsdl:portType name="oBIXSoapPort">

<wsdl:operation name="read">

<wsdl:input message="readSoapReq"/>

<wsdl:output message="readSoapRes"/>

</wsdl:operation>

<wsdl:operation name="write">

<wsdl:input message="writeSoapReq"/>

<wsdl:output message="writeSoapRes"/>

</wsdl:operation>

<wsdl:operation name="invoke">

<wsdl:input message="invokeSoapReq"/>

<wsdl:output message="invokeSoapRes"/>

</wsdl:operation>

</wsdl:portType>

<wsdl:binding name="oBIXSoapBinding" type="oBIXSoapPort">

<soap:binding style="document"

transport="http://schemas.xmlsoap.org/soap/http"/>

<wsdl:operation name="read">

<soap:operation soapAction="http://obix.org/ns/wsdl/1.0/read"

style="document"/>

<wsdl:input>

<soap:body use="literal"/>

</wsdl:input>

<wsdl:output>

<soap:body use="literal"/>

</wsdl:output>

</wsdl:operation>

<wsdl:operation name="write">

<soap:operation soapAction="http://obix.org/ns/wsdl/1.0/write"

style="document"/>

<wsdl:input>

<soap:body use="literal"/>

</wsdl:input>

<wsdl:output>

<soap:body use="literal"/>

</wsdl:output>

</wsdl:operation>

<wsdl:operation name="invoke">

<soap:operation soapAction="http://obix.org/ns/wsdl/1.0/invoke"

style="document"/>

<wsdl:input>

<soap:body use="literal"/>

</wsdl:input>

<wsdl:output>

<soap:body use="literal"/>

</wsdl:output>

</wsdl:operation>

</wsdl:binding>

</wsdl:definitions>

Appendix A. Revision History

Rev	Date	By Whom	What
wd-0.1	14 Jan 03	Brian Frank	Initial version
wd-0.2	22 Jan 03	Brian Frank
wd-0.3	30 Aug 04	Brian Frank	Move to Oasis, SysService
wd-0.4	2 Sep 04	Brian Frank	Status
wd-0.5	12 Oct 04	Brian Frank	Namespaces, Writes, Poll
wd-0.6	2 Dec 04	Brian Frank	Incorporate schema comments
wd-0.7	17 Mar 05	Brian Frank	URI, REST, Prototypes, History
wd-0.8	19 Dec 05	Brian Frank	Contracts, Ops
wd-0.9	8 Feb 06	Brian Frank	Watches, Alarming, Bindings
wd-0.10	13 Mar 06	Brian Frank	Overview, XML, clarifications
wd-0.11	20 Apr 06	Brian Frank	10.1 sections, ack, min/max
wd-0.11.1	28 Apr 06	Aaron Hanson	WSDL Corrections
wd-0.12	22 May 06	Brian Frank	Status, feeds, no deltas
wd-0.12.1	29 Jun 06	Brian Frank	Schema, stdlib corrections
obix-1.0-cd-02	30 Jun 06	Aaron Hansen	OASIS document format compliance.
obix-1.0-cs-01	18 Oct 06	Brian Frank	Public review comments