Consider adding extensions popular among other distributions

[yaronkoren]

One big resource that exists now, which didn't exist when Canasta was first being developed, is the "Used by" template on mediawiki.org (and its associated data file), which show, for every extension and skin, which of the big MediaWiki distributions and wiki farms use it. This provides an easy way to see what the popular extensions are - and what the possible holes in functionality are for any specific distribution or farm.

I ran a script that goes through the data JSON file and finds all extensions that are included by at least four distributions/farms, but not by Canasta. The list is below - note that this does not count use by Wikimedia sites (that's a separate template), although many of these extensions are also used by Wikimedia. I put some Canasta-specific notes (mostly negative) next to some of them.

• Babel
• CLDR
• Comments - same functionality as CommentStreams and, perhaps, DiscussionTools?
• CreateRedirect
• DynamicPageList3 - functionality more-or-less matched by both Cargo and SMW
• EasyTimeline
• EmbedVideo - basically the same functionality as the (very poorly-named) "Embed Video (fork)" extension, contained in Canasta
• FlaggedRevs - similar to Approved Revs, also on its way out?
• GeoData - can be done with either Cargo or SMW
• GlobalBlocking
• GlobalCssJs
• GlobalUserPage
• MassEditRegex - same as Replace Text? Also, currently has a denial-of-service security leak
• Moderation - close enough to Approved Revs? (Also covered in Add Moderation extension #203 )
• NoTitle
• PDFEmbed - similar to PDFHandler, contained in Canasta
• PortableInfobox
• RandomSelection
• RegexFunctions
• RelatedArticles
• RSS - can sort of be done via External Data, although not nearly as simply
• TabberNeue - similar to Header Tabs
• TemplateStylesExtender
• TitleKey - only useful if CirrusSearch is not being used
• Translate
• VariablesLua - relies on the Variables extension, so might stop being useful once the Parsoid switchover happens
• YouTube - matched by both "EmbedVideo (fork)" and Widgets?

What do people think - which, if any, of these are worth adding? Certainly there's a case to be made for some of these than others. In some cases, it may boil down to a philosophical issue of how "opinionated" Canasta should be - is it harmful to users, for example, to include DPL3, even if they might be better off using Cargo or perhaps SMW?

Speaking of big changes since Canasta's beginning: one other big change is the introduction of Recommended Revisions. Canasta currently simply uses Recommended Revisions for its list of extensions and skins, but it doesn't have to: extensions could be added to Recommended Revisions, which Canasta then simply removes for its own list. So another possibility for any of these is to add them to Recommended Revisions but not Canasta.

[tosfos]

Yes, I think we need a clear philosophy for what extensions Canasta should include. Quoting from the Contents page, "they must work well with Canasta's MediaWiki version (1.43); they must be well-maintained, and be likely to remain well-maintained in the future; and they must provide important functionality that no other extension or skin that's included does". And I think that is correct for Canasta's "one-size-fits-all" approach.

It would be nice to have better documentation of Canasta philosophy like the outcomes of #205. (I hear MediaWiki is good at that kind of thing.) Speaking of which, I don't know if my "Yes" answers below should count as a "second opinion" because I haven't done research or even used some of them.

a philosophical issue of how "opinionated" Canasta should be - is it harmful to users, for example, to include DPL3, even if they might be better off using Cargo or perhaps SMW

We can't make a blanket statement that one extension will always be better in every use-case. One simple illustration is, "That extension looks amazing but we've been using this other one for decades and we have thousands of pages we'd have to convert." On the other hand, we should consider the effort required by the Canasta team to maintain/test all those extensions. So I think we should limit to the most popular extensions in use. That is not really an opinion, so we're not being "opinionated".

And yes, it does create a bit of a contradiction in goals between Canasta and Recommended Revisions. It makes sense for RR to make a recommendation for EVERY extension. Even if it's awful, unsupported or a better extension exists, it would be nice for admins to know which version works best. But at least right now it's the same people managing both projects so I'd (soft) punt this to reduce workload

With that all in mind:

Babel

I don't see this as being very useful in business wikis, but it makes sense for hobby wikis. And it fits the criteria, so yes.

CLDR

Yes.

Comments - same functionality as CommentStreams and, perhaps, DiscussionTools?

No. Too similar to CS.

CreateRedirect

Yes.

DynamicPageList3 - functionality more-or-less matched by both Cargo and SMW

It's popular. Soft yes.

EasyTimeline

Yes.

EmbedVideo - basically the same functionality as the (very poorly-named) "Embed Video (fork)" extension, contained in Canasta

No. Unmaintained.

FlaggedRevs - similar to Approved Revs, also on its way out?

No.

GeoData - can be done with either Cargo or SMW

Yes. It appears to offer many additional features.

GlobalBlocking

Will it work well on Canasta's farm? If so, yes I guess.

GlobalCssJs

No. The XSS possibility makes it not suitable for general (beginner admin) use.

GlobalUserPage

Will it work well on Canasta's farm? If so, yes I guess.

MassEditRegex - same as Replace Text? Also, currently has a denial-of-service security leak

No.

Moderation - close enough to Approved Revs? (Also covered in Add Moderation extension #203 )

Soft no.

NoTitle

Maybe. It's hard to think of a case where you would want no title instead of some other title. And you can do this with CSS.

PDFEmbed - similar to PDFHandler, contained in Canasta

Yes. This extension has a better syntax and is much more configurable.

PortableInfobox

Hard maybe.

RandomSelection

Yes.

RegexFunctions

Yes.

RelatedArticles

Yes.

RSS - can sort of be done via External Data, although not nearly as simply

Yes.

TabberNeue - similar to Header Tabs

No.

TemplateStylesExtender

Yes.

TitleKey - only useful if CirrusSearch is not being used

No. Not sure why someone wouldn't use CirrusSearch. Canasta makes it easy, right?

Translate

🫣 This meets all the Canasta criteria. Still, no.

VariablesLua - relies on the Variables extension, so might stop being useful once the Parsoid switchover happens

Yes. If it works, it works. We can worry about the future in the future.

YouTube - matched by both "EmbedVideo (fork)" and Widgets?

No.

[yaronkoren]

@tosfos - thank you for that detailed response. I may have kicked a hornet's nest by bringing up philosophy. :) It's true that Canasta has a well-defined philosophy already - and it's also true that ultimately, decisions still have to be made on a case-by-case basis.

The philosophy of "Recommended Revisions" is another tricky one. Perhaps in theory it should indeed hold a revision for every extension; or perhaps it should be highly opinionated; or maybe something in the middle. For now, it holds exactly the same extension and skins as Canasta, so its philosophy is de facto the same.

As for the extensions: I either already agreed, or your arguments convinced me, about these extensions: Babel, CLDR, CreateRedirect, EasyTimeline, RandomSelection, RegexFunctions, RelatedArticles, RSS, TemplateStylesExtender, VariablesLua. So that's two votes already in favor of adding these 10 extensions. (One of them, RSS, was already discussed in #192, but there were no strong arguments against or in favor of that extension at the time.)

Of the ones you said "Yes" to, here are the ones I'm not sure about:

• DynamicPageList3 - yes, it's a popular extension, but my view is that Cargo can do all of it just as well. I may be too close to the subject matter here! It would be interesting to hear if anyone else supports adding DPL.
• GeoData - requires more explanation/discussion; I don't know what this can do that Cargo and SMW can't, besides being able to define coordinates on a whole bunch of other planets and moons. (!)
• GlobalBlocking, GlobalUserPage - I assume these can be used with Canasta's wiki farm setup, but I don't know how; requires some research, or thought.
• PDFEmbed - this must have, as you say, some advantages over PdfHandler, since that one is bundled within MediaWiki and this one still remains popular; but it does seem strange to include two extensions that do the same pretty specific task. Plus I don't really understand the differences.

[yaronkoren]

The ten extensions discussed before were just added to Canasta, in #513. That leaves 17 extensions that are popular in general but may or may not be helpful within Canasta. For convenience, let me re-list them here, but now split up into two lists: those which no one has yet expressed support for, and those that I, @tosfos or someone else has argued for inclusion of, at least tepidly.

No support:

• Comments
• EmbedVideo
• FlaggedRevs
• MassEditRegex
• TabberNeue
• TitleKey
• YouTube

Some support:

• DynamicPageList3
• GeoData
• GlobalBlocking
• GlobalCssJs
• GlobalUserPage
• Moderation
• NoTitle
• PDFEmbed
• PortableInfobox
• Translate

Any additional thoughts/votes are welcome.

[labster]

On DynamicPageList3, I'm pretty sure that Cargo can do more things than DPL. However, with DPL, you just enable the extension, and don't need any extra configuration, which means that you can explore data without structuring it beforehand, or thinking about a data model at all. But it comes at expense of speed. So it's not quite the same tool. In Miraheze's context, it's more difficult to set up Cargo for people (though we still do it).

That said, at this point I would just wait for DynamicPageList4. No, I'm not joking: Universal-Omega/DynamicPageList4#348

PortableInfobox is particularly useful for migrating wikis from Fandom, if that ever comes up. That said it has the same problem as Variables/VariablesLua that it will be hit by the parser migration. It can be rewritten, it just hasn't been yet.

[yaronkoren]

@labster - thanks for the comments. It's true that DPL can be a convenient, lightweight tool. Though I didn't know about DPL4 - wow. Perhaps we should indeed wait and see...

Good to know about PortableInfobox, also. I didn't know what its standard use case was - if migrating wikis from Fandom is the standard thing, then that seems worthwhile, but I don't know if it's a common-enough need for inclusion in Canasta. The fact that it needs maintenance is also good to know.

[yaronkoren]

Actually, now I want to remove EasyTimeline. I wish I had looked more into this extension before - it's certainly useful once it's installed, but installing it is not an easy matter (no pun intended). As specified here, you also need to:

• Install Ploticus
• Install Perl, if it's not there already (Canasta doesn't have it)
• Install the FreeSans font

Also, it turns out that the actual name of this extension is "timeline" (small t)! "EasyTimeline" is the name given on mediawiki.org, but for actual downloading and installing, it's "timeline" - which means that the current download presumably doesn't work at all.

Any objections to getting rid of it? (Hopefully not.)

[yaronkoren]

Closing this now - it led to 9 extensions being added, which feels like progress. It might make sense to do a similar kind of analysis before the next big Canasta release.