[packman] PMBS - quo vadis?

[Richard Brown]

On 24 February 2017 at 10:29, Stefan Botter <jsj at jsj.dyndns.org> wrote:

> And here again is the misunderstanding. Packman is not just there to
> provide the same packages as openSUSE, only fully working ones. Packman
> is a repository to host interesting software, which is elsewhere not
> found. Or crippled. Or old.

I do not want to comment too broadly on the discussion going on here -
there's a fair bit of quotation of me already.

But I have to pick out this sentence and point something out.

Every time Packman provides a package which COULD be in the
distribution(s) it's providing the package for, it is missing an
opportunity to help the upstream distribution it's meant to be working
with.

This is not a good sustainable model for Packman, and is not what I
would consider a friendly or healthy relationship between projects.
I consider it akin to openSUSE carrying patches without submitting
them to their respective upstreams, something which is seriously
frowned upon within the openSUSE project, and I do not apologise for
carrying forward that distaste when I see it in contexts like this.

It's not good open source practice, and it's really sad to see it
anywhere, but especially in a project that is close to one I care
about.

Putting that aside, every time Packman diverges from the distribution,
either in the above example of additional packages or in the form of
changes to overlapping packages, it's introducing scope for users to
experience broken behaviour, unexpected behaviour, or even benign
usability changes, which users still have to deal with.

To do this properly, you need to have solid quality controls and
processes, robust documentation, and multiple channels to keep a broad
userbase informed what's going on in your broad-based repository.
These are standards which not only distributions but any broad-based
repository provider adheres to, not just for the sake of their users
but to reduce the maintenance workload long term.

I feel Packman has little or no capability in any of these areas. I
think this is a recipe for repeated disappointment in the eyes of your
users, and as a package provider I think failing to address these
basics of software distribution is irresponsible.

With the current size of the Packman community & the hardware
available, I do not think it's feasible to suggest that Packman
invests a large amount of time, effort, hardware into the tooling and
processes to address those problems head on. Therefore I think it
would be the responsible thing to seriously and wholeheartedly adopt
any effort to reduce the scope of Packman to something much narrower
than "Packman is a repository to host interesting software".
This would at least mean you'd be in a position to at least limit the
breadth of risk to users by limiting the number of packages that are
provided without robust review, QC, QA, and active documentation.

openSUSE would be happy to take everything we can to help reduce that
burden upon Packman, we can work together to solve these problems, we
may not share build servers but almost all of Packmans users are
openSUSE users and I want to see them get the best experience possible
from both of our projects.

Regards,
Richard