Difference between revisions of "Summer of Code 2008"
Revision as of 10:41, 2 April 2008
The XMPP Standards Foundation has once again been accepted to participate in the Google Summer of Code for 2008. We're using this page, the JDEV discussion list, and the jdev chatroom to talk about potential Summer of Code ideas. Also free free to chat with developers working on particular codebases.
XMPP is the Extensible Messaging and Presence Protocol, an XML wire protocol for real-time communication that emerged from the Jabber open-source community. Our community is not a traditional open-source project because it is not focused on a single codebase. Instead, our community is centered around open standards and open protocols. However, even though there are closed-source implementations of XMPP, we still have a strong commitment to open code and there are many free and open-source projects in our community.
We have participated in the Summer of Code since its inception and have learned many lessons, among them:
- We try to choose half a dozen excellent projects and really focus on them.
- It is difficult to choose excellent projects, so the more helpful information you can provide, the better.
- We also try to choose excellent mentors and match students with mentors.
- We expect regular reporting (via blog) and weekly or bi-weekly meetings.
- We take the Summer of Code very seriously and we expect our students to treat it like a full-time job.
This page lists some potential project ideas that students can work on. This is similar to a "Request for Proposal" process for your summer job. The mentors and other project members have defined these RFPs as a way to help structure the summer work. Many of these projects are are relevant to our overall community roadmap. A select team of longtime XMPP developers and former mentors will review all the student proposals and "interview" many of the students so that we can make the best possible choices.
If you have any questions about the XSF's involvement with the Google Summer of Code, please contact Peter Saint-Andre via email or (preferably) IM.
Thanks for your interest, and good luck!
How to Apply
Application instructions are available at the main GSoC site -- see here.
Here is what you should include:
- Some details about you (what code languages you like, what programming courses you've taken, etc.)
- Results of your preliminary research into the protocols and codebases you might work on
- Your ideas about how to approach the project, including an outline of what work you think is required, a rough timeline of the work, etc.
- Possible problems you think you may face
In general, we consider your GSoC project to be a summer job, so try to provide detailed information that will enable us to decide if you deserve to earn $4500. Our project ideas are like "RFPs" -- your application is your proposal to get the job. Please treat it seriously. Thanks!
Jingle video using iaxclient
- Proposed Mentor: Mats Bengtsson
- Programming language: C (knowledge of Tcl/Tk is a plus but not required!)
This project involves two distinct tasks:
- Write a Tcl/Tk widget integrating video from the iaxclient package. There already is a Tcl package available in Coccinella for iaxclient audio that can be reused.
- Include this video widget into Coccinella. A Tcl package supporting the Jingle protocol exists and is being used by the audio package.
Iaxclient as a cross-platform C library for unix-like systems, Windows, and Mac OS X. It is fairly high level and takes care of codecs, some negotiating, network, and grabbing. The work involves mainly C coding and some Tcl coding. Only C coding knowledge is required to participate. There already exist video widgets for Tk which can be used as templates (QuickTime and DirectShow). The student may pick any of the supported platforms for the work since all components are cross-platform.
Jingle file transfer
- Proposed Mentor: Mats Bengtsson
- Programming language: C/C++ and/or Tcl
Transporting files from one user to another is plagued by NATs and firewalls. In ancient times a standard Tcp connection was all that was needed. Instead various application level protocols have been implemented on top of Udp which handles NATs much better. This "reinventing the wheel" by having to reimplement a kind of pseudo Tcp stack on top of Udp is a very unfortunate situation which lacks standardization.
This projects aims to use Google gTalks Jingle implementation as a starting point. Ufortunately it is written in C++ and deeply tied into the application notifier system and therefore not reusable as is. Instead the task will be to use the Tcl Udp C code implementation, and essentially use gTalks PseudoTcp class as a wrapper. There are two possible options here: either do a pure Tcl script implementation of the tricks used in PseudoTcp, which I'm not sure is possible, or just translate the C++ class into C code and wrap up the tcludp package to a pseudo tcp (ptcp) command. This should work just like the standard Tcl socket command and should be able to use it transparently.
Jingle contains several parts but the actual jingle xmpp protocol stack is already running and used for voip. A limited STUN client has also been implemented in pure Tcl, but some support for the ICE protocol will likely be necessary.
Aquarium for kids
- Proposed Mentor: Mats Bengtsson
- Programming language: Tcl/Tk
This project involves the creation of an interface for kids to Coccinella. You need to create a new roster style, the fishy roster style, with an aquarium containing fishes that represents the contacts (like ralphm's Jabber Aquarium). The fishes need to be be animated with User Mood. For example, when a contact is happy, the fish will swim quickly in the aquarium with a smile on his face. The graphics will be done in 2D, but when time allows, they can be made in 3D.
The programming tools will be Tcl/Tk and the tkpath graphics extension package  which provides high quality graphics on all platforms. There exists an albeit primitive SVG importer so the actual graphics elements can be obtained from a SVG drawing tool.
This project will involve mostly new code, and few boring code rewriting.
- Proposed Mentor: Kevin Smith
- Programming language: C++ / Qt toolkit
Psi's current history system is not very usable - this project would involve a student replacing it with a more usable UI, implementing history for groupchat, and using server-side history. Since Psi is a widely-used client, this project would be of particular benefit to a large number of the Jabber/XMPP community.
Themable WebKit-based Chat Dialogs
- Proposed Mentor: Remko
- Programming language: C++ / Qt toolkit
It's a widely accepted fact that Adium is one of the best looking IM clients out there. Besides the fact that nearly all native Mac OS X apps look good, Adium draws a lot of it looks from its themable chat dialogs. These dialogs use WebKit to draw their contents. WebKit is an open source web engine, used by Apple in Safari, and even throughout the whole Mac OS X operating system. The good news is that, in the next Qt release, there will be a built-in integration with WebKit, which is good news for Psi.
In this project, the student will create a themable, WebKit-based chat dialog. Additionally, the chat dialog should be themable with the distributed Adium themes, in order to be able to enjoy a wide set of themes out of the box. Another property of this chat dialog is that it would be decoupled from Psi itself, which would ease initial development, testing, and allow easy porting to other Qt-based IM clients.
- Proposed Mentor: Yann Leboulanger
- Programming language: Python / GTK+
The goal would be to refactor the way events are handled in Gajim, in order to implement a plugin system. There need to be some GUI hooks, and data flow hooks. The objective is to setup the system and to add some hooks as a proof of concept. Some more hooks will be added later. Second objective is to write one or two plugins as examples.
- Proposed Mentor: Yann Leboulanger
- Programming language: Python
Title tells all. Bosh is an authentification methode based on HTTP. It's described in XEP-0124. It has to be implemented in our fork of xmpppy library.
Modern Web-based Jabber/XMPP Client
- Proposed Mentor: Remko
Despite all the AJAX/Web 2.0/... hypes, and despite the plethora of closed web-based IM clients (Meebo, Google Talk Gadget, GMail/Talk, ...), there still exists no open, free, and attractive web-based Jabber IM client. In this project, the student will try to fill this void, or at least build a solid foundation towards the ultimate web-based IM client.
stpeter saith: You mean like Soashable? :-)
PubSub ATOM aggregator
This is an idea to develop a XUL xmpp4moz-based PubSub ATOM aggregator. The aggregator would support XEP-0060 PubSub, ATOM-over-PubSub and XEP-0136 Message archiving so that feeds could be synched across multiple computers and their state saved. It could be possible to download all feeds for offline/disconnected mode. The aggregator could be developed as a stand-alone application and as an extension for Firefox and Thunderbird, or/and integrated in SamePlace. -- Kael
I would be willing to (help) mentor this project -- -- Itay
Mozilla PubSub application configuration extension
This is an idea to develop an extension for xmpp4moz/SamePlace to save Firefox and Thunderbird settings remotely, following the PubSub persistent storage model defined by XEP-0223, similarly to ACAP (cf. RFC2244 and ACAP Internet-Drafts).
This extension would allow to save Firefox browser settings, bookmarks, history, cookies, Sherlock and OpenSearch search engine plugins ; and Thunderbird email, newsgroups and feeds accounts, and configuration settings on an XMPP server. -- Kael
Mentor: MickaÃ«l RÃ©mond (email: email@example.com / jid: firstname.lastname@example.org)
Those proposals for GSOC projects could be implemented in any Jabber server and programming language. However, they are specially easy to implement in ejabberd for several reasons:
- ejabberd modularity allows to implement those features without requiring changes in ejabberd core
- there are several contributed modules that can serve as example for some parts of the features
- The Open Source Erlang/OTP distribution includes many libraries that implement common small tasks, and can be used freely because they are available in any ejabberd deployment
- The erlang declarative programming language allows easy prototyping: no need to declare functions, variables, etc just to implement a function/module and try it.
Page for Account Registration & Password Recovery
Develop a module for ejabberd that provides a page to create account, change password and recover lost password.
There are two different problems that are of great importance for public servers and the XMPP federation as Jabber gets increasing popularity:
- In-band account registration of Jabber accounts (XEP-0077) can be abused easily by spammers. For example, a spammer could create a million of accounts in jabber.org easily, and then put 100 zombie machines to login to those accounts and send spam to everywhere in the XMPP Federation.
- There isn't a generalized way to recover a forgotten password of a Jabber account. Since In-band account registration (XEP-0077) is the standard way to create Jabber accounts since 1999 until now, Jabber accounts do not have email accounts associated. Hence, it isn't possible to recover a forgotten password automatically. The solution right now is to contact an admin in the Jabber server, which will change the password manually and send an email with the new password.
This proposal attempts to fix both problems. It consits in three related tasks:
- Provide a registration page. The page requests: username, password, repeat-password, email address. When filled the form, it registers the account in the local ejabbed server. The email address is NOT stored in the user's vcard; it is stored in a secure a private way, just like the password. Optional elements:
- The page could also include a captcha to require registration
- The page could send an email instead of registering the account. If the email is responded correctly, then the account is created
- Provide a page to change the password and email addresProvide also. Of course, this page will require the valid username and password in order to change the information.
- Provide a page to recover the forgotten password. The page only requires a valid username. Then it sends an email to the address stored for that account. If the email is answered correctly, the password is changed and the new one is sent by email.
- ejabberd 2.0.0 allows to provide web pages easily using the feature request_handlers in the service ejabberd_http. Example: mod_http_bind, mod_http_fileserver, mod_webpresence.
- There is an old and experimental contributed module, mod_passrecover, that implements part of task 3. It could be used as an example
- Example of registration page, this one is written in PHP: http://www.jabberes.org/jrt/
Implement an ejabberd module that tracks how much time a Jabber account is online in each presence type. Reports can be check by the user and/or admin.
Its usage is voluntary for the user, as programs like KArm, GnoTime, TaskCoach. Instead of requiring a specific desktop program, this project uses the presence set in the Jabber client to track the time invested in each task.
- When I start my work in the morning, I start my Jabber client and login to my working account. Then set presence 'busy' with status message 'Doing GSOC: write cases for test suite'. The Time Tracker captures that and stores the meaningful information.
- At midday I change to 'xa' with text 'Launch time'.
- During the afternoon I set to 'busy' again, but status message 'Doing study: homework'.
- When I am interrupted for more than 5 minutes, my Jabber client sets to 'autoaway', and the Time Tracker will also notice that.
At the end of the day, I can access the Time Tracker interface (web, adhoc or whatever) and ask the report. For example, it can report:
- busy - Doing GSOC: write cases for test suite - 5 hours and 30 minutes
- busy - Doing study: homework - 2 hours and 10 minutes
- xa - Launch time - 40 minutes
Last Week (last 7 days):
- busy - Doing GSOC: write cases for test suite - 30 hours and 30 minutes
- busy - Doing study: homework - 7 hours and 10 minutes
- xa - Launch time - 2 hours and 53 minutes
- xa - Dinner time - 2 hours and 12 minutes
- online - Browsing the web - 3 hours
- ejabberd 2.0.0 provides the feature request_handlers in the service ejabberd_http, so it's easy to provide web pages. Example: mod_http_bind, mod_http_fileserver, mod_webpresence.
- There are contributed modules that track presence changes and can be used as examples: mod_logxml, mod_statsdx, ...
XMPP to XMPP gateway
With the release of exmpp, writing an XMPP client is now easy. Writing a gateway that allow user of a given ejabberd server to reuse their other or old XMPP/Jabber account it a useful project.
The development can be developed in Erlang and leverage the existing work on exmpp. It might also involve to improve the Erlang XMPP library itself.
Here are some nice features that would be nice to have in Openfire. We are still working on the list which should be ready for early next week (March 25th). If you have other ideas not listed here feel free to send us an email to discuss it or post it here.
* File Repository and Sharing (including WebDAV) * File transfers for gateways * Group Chat for Gateways * Update and improve BOSH support * Complete XMPP RFC implementation in XIFF and work on web client
For more details, visit the Summer of Code 2008 Projects.
Mentor: Gaston Dombiak (email: email@example.com / jid: firstname.lastname@example.org).
The PJS project aims to create a reliable, correct and highly scalable XMPP server framework in Python, not unlike what ejabberd is to Erlang. It was started as an undergrad project at the University of Toronto in February 2008. By the start of GSOC, the server should have the following components:
- basic connection handling
- internal message routing (with chained handlers)
- a way to run blocking handlers in threads
- presence, subscriptions, roster get/set, message
- sqlite backend for user, roster and offline messages storage
The server will not be production quality by the end of the semester, though it will have substantial progress toward this goal. Some of the things that may need to be written:
- S2S handling (already has parts of it)
- Privacy lists (may have parts of it by May)
- Clustering. This could use something like pylinda or PyRO to distribute jobs among machines.
- Pick any XEP you'd like to implement. Dialback (XEP-0220) may be already implemented by May.
Because the project is a part of a course, its code cannot be released until the course is over (around May 1, 2008). The project will be licensed under the PFL (same license as that for Python).
The architecture is as follows:
- asynchronous event handling using asyncore (low per-connection memory overhead)
- chained handlers in pairs (handler + error handler), where each can be run either in-process or in a threadpool.
- everything is a plugin (which is just a handler class)
- everything can be redefined at run-time (little protection, but lots of power)
Mentor: David Janes (email: email@example.com / jid: firstname.lastname@example.org) . Current developer: Dmitri Vassilenko (email: email@example.com / jid: firstname.lastname@example.org)
Also see Python's SOC page (under Dr.Project)
This idea is about writing a XMPP Component which acts as a bridge between a IRC network and open and federated XMPP servers. For an IRC network it looks like another server in the network which is connected via one of the custom IRC server-to-server protocols. The component should be written extensible so it's easy to make it work with the different servers like Hyperion or UnrealIRCD. For the XMPP world the component is connected to a XMPP server und some domain like myirc.example.com. When a user from XMPP world joins the Multi User Chat email@example.com, the component will map this as a login to #hello on some IRC network. Since most IRC servers are written in C or C++ it seems to be best that this component is written in C or C++ as well.
Libraries you can use are found under http://jabber.org/
Write a XMPP transport/gateway which enables to use other jabber accounts through only one. It should cover things like transporting presence, messages, roster groups, PEP events and if there is time left even more. Transports like that will make migration XMPP accounts easier.
Sync mode for clients
Write a protocol extension to create a "sync mode" for clients. The idea is to broadcast all a user's incoming and outgoing messages to all of the user's connected resources, so that the user can move between clients on different devices with no interruption. It should also have the option of preserving a client's window state, so that opening or closing a window on one client causes the window to be open or closed on all connected clients.
Java Message Service (JMS) provides enterprise messaging features that are similar to both XMPP pubsub and standard XMPP messaging with http://www.xmpp.org/extensions/xep-0184.html receipts]. This project would involve writing a bridge between JMS and XMPP so that messages could be shared across systems. This bridge might take the form of a server-side component or a common API. For more details, see here and . Possible mentor: Fabio Forno.
Universal Transport via libpurple
The libpurple library is used in Pidgin and Adium for their multi-protocol support. This component would interface from XMPP to libpurple via an existing C or C++ XMPP library such as gloox, iksemel, loudmouth, or libstrophe. As a result, the component would automatically translate to any of the protocols that libpurple supports, including AIM, Bonjour, Gadu-Gadu, Groupwise, ICQ, IRC, MSN, MySpaceIM, QQ, SILC, SIMPLE, Sametime, XMPP, Yahoo!, and Zephyr. Possible mentor: Andreas Monitzer
Serverless Messaging aka Local-Link
Write up a platform independent implementation of XEP-0174 for gloox. Platforms supported should be *NIX (including Mac OS X), linux and windows.
ISS (Instant Syndicating Standards)
PubSub Wordpress and Drupal plugins
This is an idea to develop a PubSub plugin for WordPress and Drupal with Atom notifications The goal is to ease the publication with PubSub and to create an XMPP PubSub ecosystem.
Blogs could support an autodiscovery element like :
<link rel='alternate' type='xxxx' href='xmpp:firstname.lastname@example.org?;node=blog'/>
(The "type" is yet to be determined...)
Each new post could generate a new leaf node which would contain the post per se and the comments. In this manner, users could subscribe to the thread of their choice directly from the blog by clicking on an RSS-like button.
Or there would be one node (not a separate node for each blog post!), as described at http://mail.jabber.org/pipermail/jdev/2008-March/026352.html
XEP-0070 for the Drupal Jabber authentication module
Drupal implements a Distributed Authentication System which allows to use Jabber credentials to log onto every Drupal sites (e.g. the Ejabberd site). But it requires to pass the password to third-party sites often without encryption.
Inspired from The South African XMPP Federation OpenID Server, the Drupal Jabber authentication module could support XEP-0070: Verifying HTTP Requests via XMPP so that users could authenticate to Drupal sites with their JID in a secured manner thanks to a confirmation ticket.
SSH authentication validation with XEP-0070
The idea is to implement an additional security layer for SSH authentication with XEP-0070: Verifying HTTP Requests via XMPP.
When authenticating to the SSH server by password or by key, the user would receive an XMPP confirmation message which would need to be validated to achieve authentication. In this manner, even if the SSH password or the key were compromised they still could not be used.
It has the drawback of requiring to have an XMPP server always running and may not be suitable for some cases.
XMPP Webmin interface
This is an idea to implement an XMPP interface for Webmin, a web-based interface for system administration for Unix. Inspired by the example of XEP-0050: Ad-Hoc Commands, this software would allow system administrators to remotely manage and configure their server through data-forms.
It would also allow to select events to be notified of (via messages or PubSub), and processes to monitor (via presence or PubSub), perhaps in conjunction with SNMP.
This would require a high-level of security, especially end-to-end encryption.
Additionally, it could allow to use Jabber chat windows as shell consoles.
It would be really nice to have a bot for the Flyspray bug tracking system which allows to add, delete and edit bugs via ad-hoc commands and chat messages. Additionally it could send out the XMPP notifications which is currently done by a php script.
Client-Independent Serverless IM component
- Proposed Mentor: Remko
Currently, only some clients (iChat, Gajim) support server-less chatting, a feature which is very interesting to be able to discover and communicate with people on a local network, without the need to be connected to the internet. The reason behind the slow adoption of this protocol is probably that it breaks with the very fundamental way Jabber is designed (client-server architecture), and as such is a significant amount of work for an existing client to adapt its codebase. The aim of this project is to bring server-less messaging to every XMPP client, by implementing it outside of the client.
The goal of this project is to implement a lightweight Jabber daemon that acts as a regular jabber server on the client side, but uses Link Local Messaging on the server side. By connecting your favorite client to this daemon running on your own machine, you enable server-less messaging for this client. The project should be written in a cross-platform way to reach a large population of clients. Language can be chosen, as long as there is a decent library available to do zeroconf networking, and preferably a (core) xmpp library as well (examples include Python and C++).
Some information about link-local chatting (biased towards Psi) can be found here.
Client-Independent D-Bus Service for Jingle Audio
Many Jabber clients will want to implement the Jingle Audio protocol (and eventually other Jingle protocols as well), but may lack the technical possibility to keep an internal implementation. (or just the will to reinvent the wheel â˜º) Tox (Talk Over XMPP) is a D-Bus service meant to provide Jingle functionality to a Jabber client able to communicate over D-Bus. It uses the Farsight library.
The goals of this project are:
- Beat Tox into good enough shape to perform audio conversations
- Make at least one Jabber client implement the Jingle protocol through Tox (possibly jabber.el, which is the current development platform)
- (optional) In said client(s), support the protocol used by the Google Talk client for audio conversations
- (optional) Jingle Video
- (optional) Jingle File Transfer
This project needs a mentor, as User:Legoscia will be offline during great parts of the summer.