If making self-links clickable is a deliberate new feature rather than a bug, it should follow standard link behaviour:

1. The cursor should change to a hand pointer instead of an I-beam one, as seen here:
   
   It does so correctly only in Minerva:
   
2. The self-link should receive an underline on hover (text-decoration: underline;). For comparison, standard links correctly show that underline:
   
3. The text colour should get blue on hover consistently. Currently, this behaviour is inconsistent. When I'm logged into my account on Opera GX, the link turns blue on hover across Vector 2022, Timeless and Minerva:
   
   However, it fails to do so on other browsers, skins or logged-out sessions:
   

This issue was noticed and briefly discussed on-wiki.

The proposal at T69486#7613226 was to use CSS to disable the link behavior. However, as far as I'm aware this isn't actually possible. You can use pointer-events: none in the CSS to disable clicking and hovering, but that doesn't disable a screen reader or keyboard user from tabbing to the link and pressing enter. Parsoid could be changed to add tabindex="-1" aria-disabled="true" to the html tag itself if we really have to keep the href element, but it seems like just adding the target as a dataset property for DiscussionTools to read as opposed to an href would be less of a hack.

DiscussionTools has been able to handle self-links without href generated by the old parser, so it could surely be adapted to handle them when generated by Parsoid as well.

For Parsoid the library, this is deliberate behavior, because the original idea is that a link is a link, and self-links are reading-adapted behavior. So, if one were to take a Parsoid fragment from a page and embed it elsewhere, the "self-link" is no longer a self-link. So, Parsoid's output should be agnostic. And, the DiscussionTools scenario is one such (while recognizing that DT can be adapted to work with the removal as Bartosz notes above).

However, we now have a model where there is a separate OutputTransformPipeline that adapts the canonical parsoid output to different contexts, and desktop readviews is the first pipeline we have for this. And, in the future, we expect PCS and MobileFrontend (which are right now their own extensions / services) to effectively be their own separate Transform Pipelines that adapt Parsoid's canonical output.

All that said, now that we have the OutputTransformPipeline setup, we should be able to include a transform there to suppress self-links altogether if that is what is desired per legacy parser behavior. And, depending on where in the pipeline we put this self-link removal transform, DiscussionTools shouldn't be affected and it can still see the original canonical Parsoid HTML and shouldn't be impacted.

Should I open new tickets for the unstable behaviour shown in T424777#12016849 then?

If we make the change I am proposing above, then self-links will not be clickable, and so you shouldn't need those new tickets. So, maybe hold off for a bit?

Ok, thanks!