Re: PL support

[Dmitry Gutov]

On 09.05.2020 21:19, João Távora wrote:

> Not many immediate "killer" advantages, Yuan Fu, but:
>
> - eglot.el would be simplified, tho maybe only slightly.  That is good.

Splitting a program would simplify it? I'm somewhat skeptical so far.

> - language specific quirks (that do exist despite LSP) would be dealt
>    with in the corresponding mode, not Eglot, by using Eglot's
>    existing interfaces.

That sounds fine, but then you'd have to convince major mode authors to 
set these settings. And educate them on what values they should use.

Considering you likely know more about C/C++ LSP servers than Alan, for 
example, that doesn't sound productive. And python-mode is 
unmaintained... js-mode hasn't seen a lot of dedicated development 
either. You get my drift.

> - Eglot could grow _more_ programmatic interfaces for that
>    to happen. It doesn't have them because it's the chicken and
>    the egg.

We'll have to discuss those.

> - More importantly, many bugs that target Eglot's UI but are actually
>    Emacs's would come here. Discussing them in Github and hailing (mostly)
>    Stefan and Dmitry there works, sort of, but it would be better if we 
> used  the
>    Emacs bug tracker (yes I know there are strong opinions on this).  But at
>    the very least people like Eli and Richard would be able to participate
>    regularly in those discussions, and provide insight that just doesn't
>    reach the Github-sphere.

Umm. As a person with a fairly opinionated approach to package 
development yourself, I think you might underestimate certain downsides 
in sharing this responsibility like that.

And it doesn't look like Eli and Richard have a lot of free time to get 
into the particulars, or fix Eglot's bugs. I don't either (so far).

> Let me give you an example: didn't your eglot-box thing end up being
> an eldoc-box instead?  It should be in eldoc.el, it's a pretty good idea!
> Well if eglot was in the core, you'd just automatically do it for eldoc.el
> and  get help on how to do it from seasoned Elispers who hang around
> here and not Github.

How would that help? Eldoc has a defined interface. If eglot-box could 
be based on that, it could just be considered for contribution on its own.