[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Texmacs-dev] git & texmacs
From: |
Massimiliano Gubinelli |
Subject: |
Re: [Texmacs-dev] git & texmacs |
Date: |
Wed, 19 May 2021 13:49:26 +0200 |
Dear Sebastian,
thanks for the pointer, I was not aware of this. I know Darcs and I'm very much
interested in a principled theory of patches and Rust programming.
It is not clear what advantages Darcs or any other alternative version control
system will bring to us as a whole.
I think for a project like TeXmacs which has a very small number of people
which are able to reliably manage the program as a whole it make sense to have
a central repository. Having a distributed system is not the main reason to
choose git (or something else).
Let me write the KEY motivations I had for suggesting a switch to git:
1) git has a notion of committer and author of a patch (possibly this is true
in other systems). I find this crucial for community-building since everybody
can contribute directly to the code via pull request to the maintaners, their
contribution is clearly marked. I'm not aware that this is available in svn and
to me is a serious problem because do not encourages in the correct way the
contribution of the community.
2) git has github and gitlab which are established platforms to develop
communities around open source software.
Personally I manage several git branches of TeXmacs, and it seems easy to me.
Note that I'm not a git power user, essentially I know 4 commands and it rarely
happens to me to have complicate merge or dependency graphs. I do all the work
in one branch and then just commit the whole to svn.
The ability to do correct merge is not fundamental to me, we rarely have
complicated merges and we do not have a very complex tree of development
branches.
git is used already in TeXmacs by many of the contributors like Darcy, me,
Miguel, Philippe. It seems easy to move that way.
Joris was somehow ok with the idea to try to manage a development branch in git
while keeping the stable branch in svn on savannah. I think this is the good
decision to evaluate the suitability of this technology for our project.
I would like to hear possible defects of git which are relevant to our
situation.
A byproduct of adopting git could be that we develop better tool to integrate
TeXmacs & git which is important for our users. The colleagues I know which
uses version control systems use git and not svn or others, so we should try to
give nice way to manage TeXmacs documents in git.
More sophisticate revision systems like Darcs or Pijul could be interesting to
manage conflicts in online collaborative editing of documents. I still have to
study the way it is implemented this now in TeXmacs.
Best,
Max
> On 19. May 2021, at 12:18, Sebastian Miele <sebastian.miele@gmail.com> wrote:
>
> Hi Massimiliano,
>
> Massimiliano Gubinelli <m.gubinelli@gmail.com> writes:
>> I've just found the following blog post on OCAML development which in
>> part discuss the effect of their transition to github:
>>
>> https://discuss.ocaml.org/t/analyzing-contributions-to-the-ocaml-compiler-and-all-opam-packages/7854
>>
>> it adds some background to our discussion of moving some of the
>> development of TeXmacs to a git-based workflow, possibly on github.
>
> There may be an even cooler kid in town than all other current version
> control system (including Git and Darcs): Pijul (https://pijul.com/), "a
> sound and fast distributed version control system based on a
> mathematical theory of asynchronous work."
>
> It is still open how far the adoption of that system will be in the
> (nearer) future. But when it does get widespread adoption, it probably
> is just superior. Not the least, because it is based on a mathematical
> theory that may be far more rewarding and fun to learn.
>
> Maybe Joris is hesitant to switch to Git, because he does not want to
> spend the time to learn it. And that I would find a very good reason.
> It is just tedious and boring, and takes a considerable amount of time
> from a valuable life. That may be completely different with Pijul. The
> time spent learning may be more of a time spend in enlightening and
> general insights about the core problem at hand. Contrary to (a lot of)
> time spent on many details of yet another half-baked solution.
>
> However, it remains to be seen how far the adoption will be in the
> (nearer) future.
>
> Best wishes
> Sebastian