20

Version 92.1 by Thomas Mortagne on 2009/08/06 20:16

english version of the things that need to be done:

We have the whole mediawiki database and the images so the best is to do the migration on our side and produce a xar we will just have to import when we are at AFP. That way we can be sure everything is ok and we can fix some things by hand before we go to AFP.

mediawiki -> xwiki xar

Upgrade mediawiki

First problem: the version of mediawiki AFP have (1.4.2) does not support export so the first thing to do will be to upgrade the mediawiki database on a more recent mediawiki version. This mean at least 1.5. I tried briefly with last version of mediawiki but i had some encoding issues.

Export mediawiki

See http://meta.wikimedia.org/wiki/Data_dumps

Import mediawiki in an empty wiki

Things to support in the importer

Links
Wiki links has to be respected of course. The syntax of mediawiki and XWiki are not the same so the links reference will have to be converted. Also sometimes links target will have to be rename for many reasons so the links have to be correctly "refactored".

MediaWiki namespaces -> XWiki spaces
They use only one level of namespace (kind of multiple space) so the best is to convert namespaces into spaces directly.

Default space

I think the importer UI should ask for a default space for mediawiki pages which does not have any namespace since in XWiki it will have to be in some space.

Spaces main pages

In MediaWiki (at least the way AFP use it) the main page of a namespace has the name of the namespace and does not have any namespace:
- main page: Wikis
- subpage: Wikis:XWiki

So we will have to differenciate pages without namespace and pages which are "main pages" of a namespace. Basically if the page name can be matched in any page namespace it's a main page.

Wikis -> Wikis.WebHome
Wikis:XWiki -> Wikis.XWiki
StandalonePage -> DefaultSPace.StandalonePage

Media files

In mediawiki there is no attachement in pages (at least in this version) all the files are stored in a central repository and linked from here in the wiki pages. We need to find an equivalent for this.

Here are some solutions:
- put all theses medias in a Media space one media per page. All medias in the same page would be too much i think since in AFP for example they have more than 130Mo of files
- attach the file to the wiki document where it's used: the client don't want it and I don't like it that much either, we sould try as much as possible to follow the source organisation.

Things not needed, what they don't use in MediaWiki

Macros
They don't use any fancy mediawiki macro wich is cool since it's a big work to support and convert all this

import of the xar by taking care of collisions

There is lots of images and the xar will be too big to import the classical way (as an attachement in the standard import UI) so we need to use the big import application made by Ludovic to do that.