Page MenuHomePhabricator

non-Wikitext system messages should be identified and converted to plain text content model
Open, Needs TriagePublicFeature

Description

Following on from the fix of [T202424]:

Some system messages in the MediaWiki: namespace are wikitext, and accept wikitext features like links and images. Their content model is Wikitext. That is correct.

Other system messages in the MediaWiki: namespace are plain text messages. If you add wikitext to them, it just shows up as code. Their content models are inaccurate; they should be plain text, rather than wikitext.

A good example is MediaWiki:Titlewhitelist.

Some of the others are very short system messages that just say "submit" or "edit" or something. Most of the longer messages tend to be Wikitext, but it's all a bit inconsistent due to the way that the software has been written.

Two steps here:

  1. Identify non-wikitext system messages
  2. These should be a plain text content model by default on new installations
  3. Existing wikis may need something to convert them to the correct format.

I appreciate that this may not be a priority. Perhaps the obvious ones should be done first, and this left as an ongoing open task.

Event Timeline

Aklapper changed the subtype of this task from "Bug Report" to "Feature Request".Oct 19 2024, 2:22 PM

not a bug

Tgr subscribed.

Not particularly related to ResourceLoader.

There isn't such a thing as a "non-wikitext system message" today. The same system message can, at different places, be formatted as plain text or wikitext or HTML-escaped. Being able to flag system messages as "can contain wikitext" or "can contain HTML" would be very useful for security; it would require something like T344180: Implement qqf.json (message format).

Non-wikitext messages would seem to fall under

  1. Plain text
  2. HTML
  3. Bizarre legacy code

If some system messages can be parsed as Wikitext in some situations and plain text in others, that's less than ideal, I suggest.

If some system messages are Escaped-HTML, then Escaped-HTML should be a page content model option, I also suggest.

Or is there any reason why escaped-HTML can't be just displayed as plain text?