[PureOS] Laniakea Python version is now online!

[Matthias Klumpp]

Am Mo., 20. Mai 2019 um 19:44 Uhr schrieb Chris Lamb <chris.lamb at puri.sm>:
>
> Matthias Klumpp wrote:
>
> > > > Since last week, all of PureOS infrastructure is now switched to the
> > > > latest, mostly Python-based version of Laniakea[1].
> > >
> > > This is great news. Congratulations Matthias!
>
> Indeed, congratulations and thank you for posting about it here too.
> My apologies for this late reply getting lost between various cracks
> in my inbox. :(

Don't worry, there is no expiry date on replies ;-)

> > > If so, perhaps we use a separate Laniakea channel for Matrix so as
> > > not to flood the PureOS channel.
>
> I completely +1 the idea of a #pureos-changes channel.
>
> I use this pattern in many projects; having that "soft" visibility is
> extremely helpful for all sorts of reasons (both technical and non-
> technical) in terms of encouraging more of a team feel, avoiding any
> feeling of being isolated from others. Indeed, I even wrote the
> #debian-devel- changes bot back in 2008 for these very reasons and it
> has been remarkably successful in these regards.

Yes, we had this in Tanglu as well and it was really useful :-)

> Can I thus resubmit my request for this to be re-prioritised? :)

It's definitely a high priority, but giving people access to easy
information about the archive is even more important I would say. That
being said, software.pureos.net is available again now, so you can
view build logs easily again. Next I will also re-add the debcheck and
migration information to the package details pages, as that's
important as well for diagnosing problems. Then maybe also refine the
pages to sort by suite as well, and also improve the
application/component search a bit (as that's the neat thing our users
may be interested to use, and it's easy to do).

Than I would actually be moving on to reinstate the messaging system
between Laniakea components, however I don't feel at ease doing that
without a proper testing framework in place, so that also has a
slightly higher priority. But then, the next thing definitely is the
message bus, Matrix bot and possibly Fedmsg bridge.

> Oh, one adjunct question; how can we contribute/learn more about this
> codebase? Whilst justifiable for many reasons, as it is on Github it
> definitely feels a little separate from the rest of the PureOS code.
> Can we ameliorate this somehow? I don't have any concrete suggestions
> at this time, alas.

On why it's on a different platform: That is simply because the
project started outside of PureOS before I joined Purism, and was for
a while shared between two derivatives. Nowadays I actually prefer it
to be not as tightly coupled to the PureOS project, because I hope
that as soon as it's mature enough more derivatives may consider using
it. That's very often easier when the infrastructure for a project is
at a more neutral place. The README file of the project as well as
some PureOS wiki pages definitely tell people who want to know who was
responsible for moving the project forward though.

On contributing: There is outdated documentation on
https://lkorigin.github.io/ which I should update soon to reflect what
the Python version has to offer (and to elaborate on some points). The
code is also documented, but generating some API documentation from it
would also be helpful, I guess.
I also gave a talk about the project at Debconf, but that's probably
not a good starting point. The talk slides however may be useful, I'd
have to modernize them as well though, just like the rest of the
documentation.
To build the project itself, one can use - rather unusual for a Python
project - Meson. Meson/ninja will build the binary parts and then
install all pieces into the right location, then one can use the
`lk-admin` tool to initialize the database and configure the system.

Currently, to make Laniakea really useful, one needs to combine it
with either dak or reprepro for archive management.
I'll certainly answer any questions anyone trying to understand any of
the components may have, and hopefully also write a nice quickstart
guide.

The problem to get all the tasks done is - as always - time ;-)

Cheers,
    Matthias