Mediawiki DumpHTML extension improvement
- This page is outdated! DumpHTML is no longer being worked on.
The Mediawiki DumpHTML extension improvement was an effort which needs to be granted for being able to provide an efficient and a handful solution to make ZIM files from Mediawiki. With the solution we propose to develop, everyone from the Mediawiki Administrator to the normal user will be able to generate best quality small or big ZIM files.
The ZIM format was choosen by the top actors around Mediawiki to provide a offline usable versions of the content. Kiwix is the reference reader supporting it, but ZIM is an open format, and they are also other readers.
The ZIM format was designed to deal efficiently with hugh amount of data, that means that you may deal with millions of pictures and text extremly quickly also on a small device like a smartphone. This format is complementary to the EPUB which is more though for small content and unable to scale.
Unfortunately, the development of the format is brake by a lack of softwares. We suffer from a lack of tools to build ZIM files and actually only a few people have the mandatory know-how and software solutions to do it:
- Kiwix, which use a solution based on a hacked version of Mediawiki DumpHTML extension, with additional custom scripts. This is currently the only one project generating (and more or less able to generate) big ZIM files from WMF projects. You may download these ZIM files here. This solution is currently not usable at all project external people.
- Mediawiki Collection extension developed by Pediapress (which is deployed on Wikipedia) is user friendly but suffer of many issues: (1) Really complicated to install on a separate instance (2) slow and not able at all to deal with huge amount of data (4) Rendering quality far away of online version (5) technical approach makes they are not able at all to tune the content rendering for offline usage.
Our years long experience showed us that the Mediawiki DumpHTML extension (at least the approach) is the best solution to export the Mediawiki dynamic generated HTML to an offline usable format. Unfortunately, they are few pain points:
- Not maintained, bugs are not fixed, new features are not implemented
- Only available for Mediawiki system admin
- Does not generate ZIM files (only static HTML and Media files)
Consequently, almost nobody uses it right now to generate ZIM files, this is too complicated and buggy. This, although a lot of people want to do that and contact the Kiwix dev. Team to help them to make a ZIM of their own content.
With this project we to offer:
- A way for everyone to generate ZIM files with Mediawiki
- the simpliest solution to use for end user
- the simpliest solution to use and deploy for Mediawiki administrators
- the solution offering the best HTML rendering quality
- the best designed solution simply to maintain and easy to improve (for example to add support to new HTML based format, like EPUB)
- the most performat solution in term of generating speed
1 - phpzim creation
This is about the creation of a ZIM [devzone.zend.com/303/extension-writing-part-i-introduction-to-php-and-zend/ PHP extension] called phpzim. phpzim is an extension allowing PHP developers to read/write ZIM files. It's based on the zimlib like pyzim, the Python extension to deal with ZIM files. phpzim is essential to:
- speed up the ZIM creation (avoiding using a postgresql database and the zimwriter binary).
- Mandatory to integration the ZIM generation directly in DumpHTML
- Essential for many CMS coded in php to generate also ZIM files
- Create a tgz of the zimlib with only the necessary for phpzim
- Create the code (c++) of the phpzim PHP extension using the GNU tools for the compilation
- phpzim should offer a easy API to read/write ZIM files with all the necessary options
- Code of phpzim should be online developed on openZIM repository and as a tgz directly compilable
- Code usage should be documented and documentation should be automaticaly generated using doxygen or similar
- Rewrite and improve  in PHP (dealing directly with the zimlib)
- ~ 4000 euros
More details: phpzim
2 - Revamping and fixing bugs
The worth point is that the DumpHTML extension is not correctly maintained and with the time, many issues were discovered. Currently, the extension is not really usable without fixing/tweaking the Mediawiki code.
The purpose of this work package is to fix the most critical bugs to achieve that everyone having a Mediawiki would be able to simply get a HTML dumps of his content and consequently being able to easily generate a ZIM file afterwards
After the revamping, the result should be perfect in HTML similar to what dumpHTML.pl does.
- Revamping dumpHTML and fixing bugs (80-120 hours)
- ~ 4000 euros
3 - phpzim Integration
phpzim would be a new PHP extension allowing to create/write and read ZIM file directly in PHP. This would be a binding of the zimlib, like pyzim in Python. With this library done, we will be able to create ZIM file directly from the DumpHTML.
To be able to get a ZIM file, the user will have to call dumpHTML.php and specify that he wants a ZIM file output (not a HTML dump) also some other meta informations like title, creator, etc.
So the user will need a system access where the Mediawiki instance runs and also to install phpzim (should be packaged).
- updated dumpHTML (40 hours)
- ~ 1500 euros
4 - Integrating Collection and DumpHTML extensions and new features
By integrating the DumpHTML and the Collection extension we want to give to everyone the capacity to easily create small ZIMs from the Wikipedia user interface with following advantages:
- exactly the same rendering as online
- no external dependency to install for the Mediawiki admin
- rendering done by MediaWiki (as fast as online browsing)
In addition we want to implement a few additional features (see the list of deliverables).
- Collection extension and DumpHTML integration (30 hours)
- DumpHTML parallel processing (15 hours)
- Build selection based on list of titles (20 hours)
- Use new rendering for mobile to make ZIM files for mobiles (15 hours)
- Make offline skin to avoid pictures (5 hours)
- ~ 3500 euros
The realisation of the whole project would take 4 months and would be supervised by Kelson (creator and lead developer of Kiwix) and a member of WMFR. Payment would be done after validation by both supervisers of each workpackage.
If you want to know more: