Hello,
Am Sun, Apr 19, 2026 at 02:13:44PM +0200 schrieb conroj:
> Hi @enge , sorry for the 11th-hour reply. If you consider the deliberation
> period to have begun already then please ignore (but I did not see a follow-up
> message to that effect).
since there were some comments I had not started the deliberation period;
I will probably do so later today.
> Today I read this with fresh eyes after a few weeks. One section ("Removal
> candidates that are building leaf packages") stood out to me as potentially
> divisive, because it takes the perspective of one class of user (a consumer of
> applications) over another (a developer of applications). When talking about
> packages that are ends unto themselves versus packages that are just means to
> an end, that distinction will always be subjective; for a developer, a rich
> suite of libraries will fall in the first category.
Indeed the definition is not exactly scientific and a bit vague; but I
think we have to live with it and let people who suggest removals make
judgements calls (to which others can still object during the waiting
period). As an example for packages falling into the second category,
I have experienced game emulators that have been abandoned upstream,
but people still like to keep for playing old games. In the first
category, fmt-8, fmt-9, fmt-10 and fmt-11 while we also have fmt-12.
Not all examples will be so clear-cut.
> Secondly, the discussion about categories of "leaf" packages makes me suspect
> that the definition of "leaf" is less universal as it seems. If we took
> @zimoun
> 's definition here for granted ("a package that does not appear elsewhere in
> any other packages") then this GCD's discussion of "leaf packages that by
> their
> nature are used mainly as inputs to other packages" wouldn't make sense
> logically.
Indeed; the sentence used to be "that *were* used mainly", but @apteryx
has pointed out an inconsistency with a formulation elsewhere that said
"are". Indeed I think the main case are packages that were used before
as inputs, but are not anymore (something like the fmt-8 mentioned above,
hopefully at some point in time). But with the "by their nature" it
makes sense - packages that are leaves and thus not used as inputs in
Guix, but that would mainly be used as inputs in principle (and maybe
still are in channels or people's private packages).
> For the reasons above, I wonder if it makes sense to drop the language about
> how "these leaves should stay, but those leaves can go". I think that the GCD
> still stands on its own without it.
Well, but this is an important distinction and not only language - some
leaves can go, others not; if we drop this section, then this case
remains undefined, and also the opposition with the next section on
packages inside the package graph makes no sense anymore (where the
conditions for removal are much stricter).
Andreas
