BObovine3dom·Sep 30, 2020
I think the main limitation is not being able to accept key presses from browser chrome. Mozilla in principle was prepared to accept a patch that added a new API that did that [1] but people have repeatedly lost interest in developing it - the reward for getting it working being a months-long negotiation with Mozilla to get it merged probably doesn't make it appealing to many. The API would also fix some but not all instances of not being able to use Tridactyl until the page has loaded.
Recently, we have added an escape hatch where if you press ctrl-, focus returns to the page even from browser chrome. This works even on special pages like "we could not load this page" where it returns you to a page where Tridactyl can run. I've trained myself to use it instead of escape and it feels quite natural. On every platform except Windows ctrl-6
Read the rest
now takes you to the previous tab everywhere, even on pages to which we don't normally have access. Users can add their own octopus-style binds which are accessible everywhere but the criteria for what works are unfortunately quite opaque.
I personally think the biggest problems for Tridactyl are in the future. Manifest v3 proposes heavily restricting the use of `eval` which our users use a lot for writing their own little scripts. I don't think it would be insurmountable, but Tridactyl already has its own little DSL, so we would probably run afoul of the spirit of the new rules and risk getting delisted or banned. We might have to take most of the coolest features of Tridactyl out (and maybe have a contrib/ part of the source where anyone can have their own scripts merged without too much review).
The other problem is that new web extension APIs tend be gated behind a very narrow definition of "user intent" - essentially, the user clicks on your extension button or presses an octopus-style keybind (like ctrl-,) - that excludes about 99% of the intent of Tridactyl users. A recent example of this is the new "optional permissions" where an extension can ask for permissions as it needs them; Firefox requires you can convince it that you have user intent or it will not allow you to ask for more permissions. (Incidentally, the keyboard API I mentioned above would allow us to prove user intent to Firefox in some limited circumstances).
So in summary: yeah, there are limitations. Lots of them could be fixed by working with Mozilla but no one has yet been motivated enough. In practice it's not as bad as it sounds.
[1]: https://github.com/tridactyl/keyboard-api