Re: [ELPA] New package: plz

[Adam Porter]

On 5/13/22 10:09, Richard Stallman wrote:

>    > plz.el is a library to make HTTP requests using curl, akin to url.el and
>    > request.el.  It attempts to provide a better API and avoid some of the
>    > problems found in the other libraries, such as with callbacks.
>
> That's concise and clear.  Thanks.
>
> This raises a question: does plz.el potentially subsume url.el
> request.el?  Would it make sense to replace those with interfaces to
> plz.el?  Perhaps not now, but in a few months once plz.el is
> perfected?
>
> I don't know the answer, but it would be advantageous to have only
> one such package to recommend and document.

Well, as Eli pointed out, as long as plz.el requires curl, it couldn't 
replace any Elisp-only libraries.  Providing an all-Elisp backend to 
replace usage of curl is an idea I might look into, but probably not 
anytime soon.

Besides that, as I admit in the readme, plz.el is still a young project. 
 url.el and request.el are much more feature-complete and mature.  So 
far I've added the features I've needed to use it in Ement.el; I'm 
hoping that, being on ELPA now, it will be used and tested more widely, 
and the needs of other users will help drive plz.el's development.

And while I prefer plz.el's API (of course), I wouldn't presume to say 
that it's necessarily better than theirs, so I wouldn't propose plz.el 
to subsume them.

Finally, I've encountered some rare process-related bugs in Emacs when 
testing plz.el, e.g.

  https://debbugs.gnu.org/cgi/bugreport.cgi?bug=50166

So far, it only seems to manifest when running plz.el's ERT tests in 
rapid succession, not when using plz.el "in anger."  Even when Ement.el 
makes tens of requests with plz.el in rapid succession, like when 
downloading Matrix room avatar images upon initial connection, it seems 
to work perfectly.  And it's defied all of my attempts to debug the 
problem at a deeper level.

So while I don't think that problem should discourage others from using 
plz.el, and url.el and request.el can have similar issues at times 
(probably due to issues in process.c--"there be dragons," it seems), I'm 
hesitant to push it too strongly as an alternative yet.

Thanks,
Adam