Sorry about bugging the list about it, but can anyone please explain the reason for not enabling the Interlanguage extension?
See bug 15607 - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607
I believe that enabling it will be very beneficial for many projects and many people expressed their support of it. I am not saying that there are no reasons to not enable it; maybe there is a good reason, but i don't understand it. I also understand that there are many other unsolved bugs, but this one seems to have a ready and rather simple solution.
I am only sending it to raise the problem. If you know the answer, you may comment at the bug page.
Thanks in advance.
On Sun, Mar 15, 2009 at 4:10 AM, Amir E. Aharoni [email protected] wrote:
Sorry about bugging the list about it, but can anyone please explain the reason for not enabling the Interlanguage extension?
See bug 15607 - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607
In general, extensions with this status haven't been implemented because they haven't been reviewed by a highly experienced developer.
2009/3/15 Andrew Garrett [email protected]:
On Sun, Mar 15, 2009 at 4:10 AM, Amir E. Aharoni [email protected] wrote:
Sorry about bugging the list about it, but can anyone please explain the reason for not enabling the Interlanguage extension?
See bug 15607 - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607
In general, extensions with this status haven't been implemented because they haven't been reviewed by a highly experienced developer.
Brion wrote a few comments there, but i didn't understand what exactly is the problem.
Дана Saturday 14 March 2009 23:20:01 Amir E. Aharoni написа:
2009/3/15 Andrew Garrett [email protected]:
On Sun, Mar 15, 2009 at 4:10 AM, Amir E. Aharoni [email protected]
wrote:
Sorry about bugging the list about it, but can anyone please explain the reason for not enabling the Interlanguage extension?
See bug 15607 - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607
In general, extensions with this status haven't been implemented because they haven't been reviewed by a highly experienced developer.
Brion wrote a few comments there, but i didn't understand what exactly is the problem.
Seconded. Brion's comments are:
I don't quite understand what this extension is meant to accomplish, or how the workflow is envisioned. What are the user interface and performance considerations?
The lack of automatic updates seems less than ideal, as does the multiple fetching of link data over the HTTP API on every page render. Management UI by manual editing of offsite pages looks pretty ugly; what could be done to improve this?
If we're going to create a special tool, there's no point in stopping so short that we still have to rely on somebody to run client-side bots to make it work properly. We've got a site with shared databases and the ability to queue updates automatically on the server side... take advantage of it!
I don't fully understand them.
If we're going to create a special tool, there's no point in stopping so short that we still have to rely on somebody to run client-side bots to make it work properly.
I think that it is a request to start looking into ways to handle interwiki links from the inside, rather than still relying on bots to update the pages.
The interlanguage extension is a nice first step towards the inclusion in core of all the work that we are currently doing with interwiki bots. But why would we rely on, again, client-side bots to update anything? The bot work seems to be smaller with this system, agreed. But perhaps these new type of bots updates could simply be transformed into Jobs and put in the Job queue.
At least the idea is to try to find a workflow that could be included in the server as much as possible instead of saying that client-side bots will handle the update tasks.
2009/3/17 Nikola Smolenski [email protected]:
Дана Saturday 14 March 2009 23:20:01 Amir E. Aharoni написа:
2009/3/15 Andrew Garrett [email protected]:
On Sun, Mar 15, 2009 at 4:10 AM, Amir E. Aharoni [email protected]
wrote:
Sorry about bugging the list about it, but can anyone please explain the reason for not enabling the Interlanguage extension?
See bug 15607 - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607
In general, extensions with this status haven't been implemented because they haven't been reviewed by a highly experienced developer.
Brion wrote a few comments there, but i didn't understand what exactly is the problem.
Seconded. Brion's comments are:
I don't quite understand what this extension is meant to accomplish, or how the workflow is envisioned. What are the user interface and performance considerations?
The lack of automatic updates seems less than ideal, as does the multiple fetching of link data over the HTTP API on every page render. Management UI by manual editing of offsite pages looks pretty ugly; what could be done to improve this?
I don't fully understand them.
Wikitech-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wikitech-l
If you don't know what this is about, refer to:
- http://www.mail-archive.com/[email protected]/msg01907.html - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607 - http://www.mediawiki.org/wiki/Extension:Interlanguage - http://meta.wikimedia.org/wiki/A_newer_look_at_the_interwiki_link - http://strategy.wikimedia.org/wiki/Proposal:A_central_wiki_for_interlanguage...
Дана Tuesday 17 March 2009 07:01:39 Nikola Smolenski написа:
Дана Saturday 14 March 2009 23:20:01 Amir E. Aharoni написа:
2009/3/15 Andrew Garrett [email protected]:
On Sun, Mar 15, 2009 at 4:10 AM, Amir E. Aharoni [email protected]
wrote:
Sorry about bugging the list about it, but can anyone please explain the reason for not enabling the Interlanguage extension?
See bug 15607 - https://bugzilla.wikimedia.org/show_bug.cgi?id=15607
In general, extensions with this status haven't been implemented because they haven't been reviewed by a highly experienced developer.
Brion wrote a few comments there, but i didn't understand what exactly is the problem.
I talked with Brion about the extension at the 2009 developers meeting in Berlin (I'm sorry it took me so long to report on this). He considered the extension a good idea in general, and he clarified to me what did he mean in his bugzilla comments. So:
"The lack of automatic updates seems less than ideal" - we actually skipped this. One of the problems is that there is no way for the new links to automatically propagate from the central wiki to the client wikis. In order for the links to propagate one has to edit the page or purge the page cache on the client wikis for every change on the central wiki.
Currently, if the extension would be implemented, this would either not be done at all (causing a delay in the interwiki propagation) or be done by bots (which is stupid). I guess the solution is to add a hook to ArticleSaveComplete on the central wiki that would go around and purge caches (how? via API again?) on the client wikis.
"as does the multiple fetching of link data over the HTTP API on every page render" - Brion would like the links to be fetched directly from the database of the central wiki (similar to how Commons image description is fetched now), instead of via API as is the case now. This should be very easy to do. (I'd keep fetching via API in the extension in case someone would like to play with it, but fetching from the database should be the default.)
"Management UI by manual editing of offsite pages looks pretty ugly; what could be done to improve this?" - Brion thought that the interwiki management on the central wiki should be done on the client wikis.
The first idea was to create some special page for the client wikis that would enable the editors to add the links to some central database (wouldn't even have to be a wiki). But this was dropped: it wouldn't be possible to control who added a link, where, someone could simply delete all the links etc.
Long story short, the final idea was to, basically, enable editing of the central wiki through the client wiki. Editors of the client wiki could, probably on some special page, edit the list of interwikis, and these edits would affect the articles on the central wiki, be viewable in recent changes and article histories there etc.
This would be humongously complicated and I don't think it could be done in a reasonable time, if at all. So, I don't think it should be done.
Interlanguage Extension ( http://www.mediawiki.org/wiki/Extension:Interlanguage ) is a MediaWiki extension that enables keeping the interlanguage links for all the languages at a central wiki, without the need for the herd of bots that maintain the links. Introducing it to Wikimedia projects is a featured proposal on the Strategic Planning wiki ( http://strategy.wikimedia.org/wiki/Proposal:A_central_wiki_for_interlanguage... ).
I have recently fixed some of its shortcomings (see https://bugzilla.wikimedia.org/show_bug.cgi?id=15607#c41 and onward) and I believe that the extension is now ready for practical use. Could anyone interested review the code and see if it is technically good enough, and if there is anything else that needs to be done? And if it is good enough and there is nothing critical missing, what would be the steps to test it in the real world and actually deploy it?
The only things that I see remain to be done, from the programming perspective, are:
- Reading the links from a foreign database instead of via the API. Is there a "proper" way to do this?
- Updating links on dependent wikis when deleting, undeleting etc. articles on the central wiki.
- Purging the pages on dependent wikis via the API, or smarter way of purging without the API. Why does purging via the API require login, when purging the usual way does not?
- Smarter way of handling job queues, probably separating each page to a separate job or similar.
But in my opinion, real world testing could start without any of that.
2010/8/6 Nikola Smolenski [email protected]:
I have recently fixed some of its shortcomings (see https://bugzilla.wikimedia.org/show_bug.cgi?id=15607#c41 and onward) and I believe that the extension is now ready for practical use. Could anyone interested review the code and see if it is technically good enough, and if there is anything else that needs to be done? And if it is good enough and there is nothing critical missing, what would be the steps to test it in the real world and actually deploy it?
When it has passed review, I suppose we could set up a few wikis at prototype.wikimedia.org and test it there. It was originally set up for the usability initiative but it seems to be used for increasingly random testing purposes now, and we have no other proper place for this.
- Reading the links from a foreign database instead of via the API. Is there
a "proper" way to do this?
Yes. You can obtain a connection to a foreign DB with wfGetDb( DB_SLAVE, array() 'wikiID' ); . Both the wiki ID and the API URL can be obtained from the interwiki table now (that is, there are fields for them, we're not actually filling those fields yet), as part of Peter's work on the iwtransclusion branch for GSoC. How have you been obtaining API URLs anyway?
- Updating links on dependent wikis when deleting, undeleting etc. articles on
the central wiki.
This is something the iwtransclusion branch also has to address. The current plan is to do that using a global templatelinks table.
- Purging the pages on dependent wikis via the API, or smarter way of purging
without the API. Why does purging via the API require login, when purging the usual way does not?
That's a bug in the API purge module; I'll fix it later today. Again, the iwtransclusion branch also needs to take care of purging. There's no code for it yet, but I don't think it involves more than updating page_touched in the remote DB and notifying Squid.
Roan Kattouw (Catrope)
On 08/06/2010 11:51 AM, Roan Kattouw wrote:
2010/8/6 Nikola Smolenski[email protected]:
I have recently fixed some of its shortcomings (see https://bugzilla.wikimedia.org/show_bug.cgi?id=15607#c41 and onward) and I believe that the extension is now ready for practical use. Could anyone interested review the code and see if it is technically good enough, and if there is anything else that needs to be done? And if it is good enough and there is nothing critical missing, what would be the steps to test it in the real world and actually deploy it?
When it has passed review, I suppose we could set up a few wikis at prototype.wikimedia.org and test it there. It was originally set up for the usability initiative but it seems to be used for increasingly random testing purposes now, and we have no other proper place for this.
Great! :) Note that, for testing purposes, it is possible to keep the pages of the central wiki in a separate namespace or as subpages.
- Reading the links from a foreign database instead of via the API. Is there
a "proper" way to do this?
Yes. You can obtain a connection to a foreign DB with wfGetDb( DB_SLAVE, array() 'wikiID' ); . Both the wiki ID and the API URL can be obtained from the interwiki table now (that is, there are fields for them, we're not actually filling those fields yet), as part of Peter's work on the iwtransclusion branch for GSoC. How have you been obtaining API URLs anyway?
I'm using $wgInterlanguageExtensionApiUrl variable that has to be set in LocalSettings.php
- Updating links on dependent wikis when deleting, undeleting etc. articles on
the central wiki.
This is something the iwtransclusion branch also has to address. The current plan is to do that using a global templatelinks table.
Interlanguage extension was lucky in that the articles that need to be updated are exactly the ones that the interwikis link to :)