DesignXMPP analysis
Version 4 (Anonymous, 02/29/2012 12:33 pm) → Version 5/17 (Anonymous, 03/01/2012 02:16 pm)
= XMPP library candidates analysis =
[[TOC(DesignXMPP, DesignXMPP_analysis, [[TOC(DesignXMPP_analysis, depth=2)]]
In this phase the existing XMPP libraries will be analyzed and one will be chosen to be used throughout the project.
== Library requirements ==
* Written in Python (C/C++ could also be used, but a wrapper would need to be written)
* Support for XMPP server (or component, details later)
* Ability to use it with the current model used by SylkServer
* Green threads
* Callback based IO
== XMPP server vs XMPP component
The XMPP protocol was designed to be extended by entities called "components" which sit on the side of the server. A component is addressed with a DNS subdomain in the following form:
{{{
component.domain.tld
}}}
This mechanism is defined in [http://xmpp.org/extensions/xep-0114.html XEP-0114].
With this architecture in place deployment of additional features to an existing XMPP server doesn't require to modify it. The SIP-XMPP gateway can potentially be implemented as either a XMPP component or an standalone XMPP server.
== Library evaluation ==
The following aspects were considered when evaluating a given library:
* Met requirements stated above
* Is it actively maintained?
* Example use cases
* Deployments in real-world scenarios
* Perceived complexity to integrate it with SylkServer architecture
=== SleekXMPP===
* Plugin architecture, each XEP as a plugin
* Client and server component support
* Threaded model (thread based scheduler)
* Active development
=== Wokkel ===
* Plugin architecture, 'subprotocols' implementing different XEPs
* Client and server component support
* XMPP server-to-server support (s2s)
* Designed to be used with Twisted (reactor model)
* Active development
== Selected XMPP library ==
After analyzing candidate libraries '''Wokkel''' was the chosen one. Reasons:
* Implemented on top of Twisted, which makes integration with SylkServer straightforward
* Support for both component and XMPP server models, allowing for flexibility in implementation
[[TOC(DesignXMPP, DesignXMPP_analysis, [[TOC(DesignXMPP_analysis, depth=2)]]
In this phase the existing XMPP libraries will be analyzed and one will be chosen to be used throughout the project.
== Library requirements ==
* Written in Python (C/C++ could also be used, but a wrapper would need to be written)
* Support for XMPP server (or component, details later)
* Ability to use it with the current model used by SylkServer
* Green threads
* Callback based IO
== XMPP server vs XMPP component
The XMPP protocol was designed to be extended by entities called "components" which sit on the side of the server. A component is addressed with a DNS subdomain in the following form:
{{{
component.domain.tld
}}}
This mechanism is defined in [http://xmpp.org/extensions/xep-0114.html XEP-0114].
With this architecture in place deployment of additional features to an existing XMPP server doesn't require to modify it. The SIP-XMPP gateway can potentially be implemented as either a XMPP component or an standalone XMPP server.
== Library evaluation ==
The following aspects were considered when evaluating a given library:
* Met requirements stated above
* Is it actively maintained?
* Example use cases
* Deployments in real-world scenarios
* Perceived complexity to integrate it with SylkServer architecture
=== SleekXMPP===
* Plugin architecture, each XEP as a plugin
* Client and server component support
* Threaded model (thread based scheduler)
* Active development
=== Wokkel ===
* Plugin architecture, 'subprotocols' implementing different XEPs
* Client and server component support
* XMPP server-to-server support (s2s)
* Designed to be used with Twisted (reactor model)
* Active development
== Selected XMPP library ==
After analyzing candidate libraries '''Wokkel''' was the chosen one. Reasons:
* Implemented on top of Twisted, which makes integration with SylkServer straightforward
* Support for both component and XMPP server models, allowing for flexibility in implementation