On Tue, Sep 03, 2019 at 10:52:36 +1200, Holger Jahn wrote:
The AUR does not provide source code hosting, on the grounds that other places like Github, Gitlab, git.sr.ht, amd so on are doing it better. Moreover, source code hosting is a resource burden on the provider, which in our case we do not have either a business or community rationale for accepting. (Again: we provide hosting for build recipes because build recipes are something specific to Arch).
Makes sense, and I would not have done it "my way" if it weren't for the fact that my source code is a single Perl file with a mere 1400+ lines of code.
Speaking of "borderline cases", such as mine:
IMNSHO 1400+ lines of Perl (or any) code don't fall into the category of "borderline cases" at all.
One of the opt-depend packages my disklow is using, "msmtp-mta", is containing a single symlink (/usr/bin/sendmail => msmtp). Perhaps, I should have a look at its PKGBUILD (it's most likely simply calling "ln -s" in its install() section).
package(), but yes: https://git.archlinux.org/svntogit/packages.git/tree/trunk/PKGBUILD?h=packag... But what would the symbolic link be needed for?
More generally, why would you upload your project in a way that is targeted exclusively for Arch users? What if users of a different distro wanted to use your software?
Because I am going to provide the packages for them myself.
I don't think one can realistically package something for all the different distributions out there; you will inevitably end up excluding quite a number of platforms. I think this should have already been obvious enough while writing that sendmail-setup.txt, where it only explains the setup steps for Arch and CentOS, and for the others there is just: [/usr/share/doc/disklow/sendmail-setup.txt]
Please understand that there are so many different mail system setups in existence today, that it is impossible to cater for more than a few basic mail system configurations here.
Ó_ô And by not providing a distro-agnostic source for obtaining the code, this also discourages other people from stepping up and packaging it themselves; it's unnecessarily cumbersome and I kind of see it as a middle finger towards distribution packagers. And because https://{that_url_there}/repos is not browsable, people can't even know for what distributions the software has already been packaged; that's a middle finger also to the distribution *users*. And even for Arch users, reviewing the code becomes difficult, because (as already mentioned): On Mon, Sep 02, 2019 at 10:46:48 -0400, Eli Schwartz via aur-general wrote:
[…] regardless of where you host it, why would you upload it as a tarball checked into git?
Essentially, we lose all the nice VCS features, most notably the ability to conveniently track changes: we now have to obtain the tarballs for the old and the new version, extract them, and diff them manually. --Tinu