01:24 <whot> DemiMarie: one term I read a few days ago (on lwn?) was "One unit of controversy" which I think is the best approach to group them

13:15 <karenw> In wl_keyboard there's an enum for keymap type. I feel like this can never actually be changed because it would break keyboard input on existing clients?

13:16 <emersion> what do you mean?

13:16 <emersion> if we were to add a new entry, we would bump the version

13:17 <emersion> so only clients using the new version would get the new enum entry

13:17 <karenw> Good point, ignore me then.

13:17 <karenw> also an enum of '0' is defined as 'raw keycodes'. Does any server actually use that?

13:18 <emersion> wlroots sends it if the compositor doesn't set a keymap

13:20 <karenw> Yuck. How do clients normally handle that? I don't really see a way in a generic client other than printing a warning and pretending the kb doesn't actually exist.

13:24 <karenw> And what would the fd parameter be in that case? I assume I still need to close it?

13:30 <karenw> Looks like wl_roots sends a fd to /dev/null

13:31 <emersion> yeah, sounds fine to just ignore such keyboards if you need a keymap

15:14 <colinmarc> It would be cool if the protocol let you send both the physical and logical key for each press, rather than being tightly integrated with xkbcommon. That was definitely a pain point in the remote desktop case.

15:32 <kennylevinsen> colinmarc: what is sent on the wire is the physical keycode, which the client itself feeds into libxkbcommon

15:37 <kennylevinsen> it's tricky to send post-keymap symbols considering deadkeys, compose, IME, ... - it's not just a matter of sending a symbol

22:36 <DemiMarie> Should libwayland limit message sizes to 4096 bytes? Older versions of libwayland imposed that limit but it didn’t seem to be intentional.

23:08 <whot> karenw: there's a bug with the keymap type protocol because there's no negotiation, you *can* send something else in the future but no existing client is likely to know about it so there's likely some protocol needed to add some "I support those formats". but since there's nothing other than XKB on the horizon...

23:09 <whot> iow works for now, if we ever get a YKB or ZKB we will run into issues

23:34 <dottedmag> WKB :-)