On September 25, 2025 7:29:37 PM CDT, Michael Orlitzky <[email protected]>
wrote:
>On 2025-09-25 15:47:24, John H Palmieri wrote:
>> Here are some recent occurrences in Sage development:
>>
>> 1. The documentation is not built by default.
>
>Is there an issue about this? I don't build the docs so I honestly
>don't know. What I do know is that when we recently switched to meson
>for building the sage library, the docs were built unconditionally. I
>know because I had to (re)add the option to disable them:
>
> https://github.com/sagemath/sage/commit/aca4aed1192
>
>They are however enabled by default, so AFAIK they should be built,
>and the intention was never to change that.
>
>
>> 2. There has been the assertion that Conda is the recommended approach for
>> compiling from source.
>
>This has been beaten to death, e.g.
>
> * https://groups.google.com/g/sage-devel/c/AvH3xq2bCfo/m/Rrr0Chp0DAAJ
> * https://github.com/sagemath/sage/discussions/39272
>
>Ultimately, what we should recommend is what is most likely to
>work. The bottom line is that sage distribution is rotting. Speaking
>for myself, now that I'm not forced to maintain hundreds of duplicate
>packages, I don't. And from the git log on build/pkgs, I'm not
>alone. With no one maintaining the packages there, it ceases to be
>what is most likely to work. Conda is a good alternative, but I also
>get the impression that the suggestion du jour is not the point of
>your post.
>
>This issue is unique in that no one needs to do anything for the sage
>distribution to bit-rot. Packages become out-of-date and incompatible
>with newer libraries/compilers all on their own; no malice on behalf
>of any sage developers is involved. No one has decided to switch to
>Conda, it's just slowly becoming the better choice. Some may find that
>regrettable, but it's not something that can be addressed by
>fiat. Reversing the trend would involve a huge, ongoing time
>commitment, indefinitely. People are free to work on it if they want
>to, but we can't legislate it.
To elaborate, most spkgs are Python packages (and most of them are dependencies
of just a few: notebook, sphinx, pip, python3).
None of them are patched, so it's purely a duplication, manual, error prone,
and time consuming, of what one can install from PyPI or a similar source in
few commands.
Now, updating a few packages often breaks some other packages, for it's
humanely impossible for one person to do dependency resolution for 300
interconnected packages by hand.
And bitrotting they are - just today I found a few spkgs which are not Python
3.14 compatible,
while working on <https://github.com/sagemath/sage/issues/40890>
My modest proposals to fix a part of it, by allowing standard packages to be
pip-installable were voted down.
Now, may I ask, why do these people, who have voted my proposals down, do not
keep up with constantly updating the spkgs, doing the necessary reviewing of
such updates, etc?
Many of them are (almost) never seen contributing code or reviews. They at most
come here and complain that Sage is hard to install, that they object to such
and such changes, etc. Have they assumed that the extra work they demanded by
their votes will be magically done for them?
I think the only way out is to get rid of these spkgs.
We don't need a bundled python, or a bundle notebook/jupyterlab. Sage can very
well be made into a pipx package (and sagelib into a pip package).
Many non-python packages are dead wood, in the sense one can always have an
alternative from your OS or distro.
E.g. gcc/gfortran, readline, etc.
I have been working on removing such packages; you might have noticed that
zlib, bzip2, pkg-config spkgs are gone (moved to pre-requisites) in 10.7; there
is a positively reviewed PR to remove boost spkg (i.e. move it to
pre-requisites).
Dima
>
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/sage-devel/5A3F8DD2-55C7-4373-B63B-0E97A2BC46F8%40gmail.com.
