Re: [nongnu] main 74116339a8 2/3: * elpa-packages (anzu): New package

[Philip Kaludercic]

Stefan Kangas <stefan@marxist.se> writes:

> Philip Kaludercic <philipk@posteo.net> writes:
>
>> Stefan Kangas <stefankangas@gmail.com> writes:
>>
>>> branch: main
>>> commit 74116339a852129b98ff79e2eb3aa35b387aa006
>>> Author: Stefan Kangas <stefan@marxist.se>
>>> Commit: Stefan Kangas <stefan@marxist.se>
>>>
>>>     * elpa-packages (anzu): New package
>>
>> Hasn't anzu been obsoleted by isearch-lazy-count?
>
> I don't know; I don't use isearch these days.  I remember using anzu
> when I did.
>
> I add packages mostly based on our formal rules, not on if I think they
> are useful or good.  Some notable exceptions to this are s.el and f.el,
> where we have decided that it is strongly undesirable to promote such
> short package prefixes.  I have also skipped some other packages that
> are clearly obsolete/no longer developed/bad or low effort, etc.

I think this might be worth discussing.  The impression I get is that a
lot of packages/libraries are not being used because they are better,
but because they are still recommended on outdated blog posts and fora
(think of linum-mode vs. display-line-numbers-mode).  More on this
below.

> I am currently focusing on highly popular packages, and anzu clearly fit
> the bill with over 1 million downloads on MELPA.org.  It also seems like
> it is still maintained, with a new maintainer since March 2020, and
> commits as late as October 2021.

The "danger" is that MELPA might give the wrong impression of
popularity: Their download counter is an absolute number, and does not
account for the age of a package or the number of downloads that have
been updates (this skews towards older packages and packages with
frequent updates), or when a package was updated (this can be manually
approximated for some packages via archive.org).

> But feel free to tell me why I'm wrong; I'm not married to it and if you
> think we should remove it again, I don't think I will have any
> objections.

This seems to be more of an general, open question on the direction that
NonGNU ELPA should head:

Do we want to collect as many packages as possible, even if the
implementations and practices are sub-optimal, are displaced by
alternative implementations in Emacs or ELPA, etc. or should we try to
restrict the packages to popular, "good citizens" of the Emacs package
space, in an effort to raise the standards and clean up "obsolete" and
"redundant" packages.  It is probably clear that I have an inclination
towards the latter position: Going forward it seems preferable to have
as many useful and idiomatic packages available directly via the ELPAs,
without burdening newcomers with duplicate functionalities.  My
motivation in contributing to NonGNU ELPA is to further this goal.

>From my restricted understanding of Emacs' history (mostly due to my
age), MELPA and EmacsWiki before it contributed a substantial
improvement (more packages, cleaner code, usage of libraries, ...) next
to their respective formal improvements (package.el support, a
centralised location for packages).  It seems to me that MELPA leans
towards completeness, adding anything from serious packages to fun and
jokes.  While I do understand the motivation/advantages, it appears to
fuel the "there is a package for that" mentality (parallel to Apple's
"there is an app for that"-slogan), that implies to download a package
for every little change, instead of writing or even copying some Elisp.
Given that package.el still makes it difficult to maintain your own
changes while updating packages, I understand why people falsely claim
Emacs has a "plug-in" system.

(From what I remember you were intending to present a talk on a related
topic at EmacsConf last year, right?)

> BTW, maybe this should be raised as a ticket on the anzu bug tracker?

What could they do with this information?  I guess they would disagree,
and that would be it.

-- 
        Philip Kaludercic