Re: Migration - where are we up to?
On Sun, Feb 15, 2026 at 05:52:05PM +0000, Morten Kjeldgaard wrote:
> One thing that may be holding back the wiki is that nobody knows where to put new content. It is obvious to most that the old wiki is going away, but on the wiki2025 front page it says:
>
> > Do not begin to migrate content here, it may be wiped without notice. More info.
>
> So if you are looking to write something for the Debian wiki you are definitely discouraged as it might be "wiped without notice", and writing for the old wiki doesn't seem quite as motivating.
New content goes on the current wiki. The new wiki is currently just a place
to try out ideas that could go terribly wrong (see later in this e-mail for an
example ;)
I think this is a strong argument for guaranteeing old content will be
migrated - you can add to the wiki without worrying your work will be deleted.
>
> In addition there is no organizational framework visible on the landing page so people wouldn't even know where to put new content.
>
> From the beginning I have been a proponent of starting the wiki from scratch, and porting content from the old wiki if the information is still relevant and of interest. I accept that there seems to be agreement that old content should be imported, but I also see that the process seems to be stuck at this point, and not much has happened, since the task rests on a couple of people with access and not a lot of time to invest.
Time is definitely a problem, but access isn't currently an issue. The big
time sink is identifying hundreds or thousands of special cases, then deciding
what they would look like in MediaWiki. I've put as much time as anyone into
that job, and I don't have any special wiki access. Anyone else willing to
give it a go is very welcome!
I concede this is a strong argument against guaranteeing old content -
promising to port every possible edge case creates a vast amount of work.
But how about this for an idea...
1. import the MoinMoin database into MediaWiki with the existing markup
2. get the raw HTML for every existing page with curl and `?action=content`
3. modify that HTML in specific ways we need to for the new wiki
* obfuscate e-mail addresses
* mangle any HTML MediaWiki refuses to touch
* maybe add [Category:foo] links?
4. upload a new version of every page with the modified HTML content
5. rename pages as needed by MediaWiki
* translations
* category pages
* user pages
* talk pages
* ...
5. manually fix pages as time and interest allows
* usually just means reverting to the last MM version and tweaking a bit
The original plan was to just grab some existing script, convert all the pages
and have a MediaWiki instance ready to go in short order. That didn't really
work out, and my previous message in this thread argues it's impossible in
principle - MediaWiki markup operates at a higher level of abstraction than
MoinMoin markup, so any good migration will involve porting a lot of pages by
hand. The plan above is kinda the natural conclusion of that argument - do the
migration at the common level of abstraction, then have people add the bits the
computer can't.
Reply to: