Re: [gentoo-project] Corporate affiliations of Council members

[Patrick Lauer]

On 2020-06-23 22:50, Andreas K. Hüttel wrote:
> Am Dienstag, 23. Juni 2020, 20:48:04 EEST schrieb Patrick Lauer:
>
>> (And this is why I'm against things like the current py2 purge: There is
>> code out there that works, can't be rewritten to py3 in a reasonable
>> time*, and hasn't been rewritten in another language yet. There is no
>
>> tl;dr: I want to be lazy, so stop breaking stuff ;)
>
> You gotta be kidding me.
>
> If your employer wants to work on obsolete (yes) stuff, he should task you
> with maintaining it.
>
> In the context of a distribution that does *not* only mean "keep things until
> they are so rotten you can't distinguish them from the floor anymore", but it
> also means fixing bugs and keeping the dependency tree in a sane state.
>
> So, in this case I would strongly recommend to the higher-ups of A***** to pay
> you (or anyone else) to commit to keeping
> * not just what *you* need in Python2 maintained and working, but
> * to maintain the whole python-related package tree in a consistent state
> then.
> Not just complaining that others don't volunteer the work.
>
> (This argument applies equally to other cases, like S***.)
>

So, uhm.
I'll let you claim it was late and the beer was talking.

But that's a totally incoherent hallucination you're projecting on me.


The problem with 'obsolete' stuff is that it's often very useful, but no 
one wants to spend time rewriting it.

So for example, codeswarm. The log parser it uses to convert 
cvs/svn/git... logs into an intermediate representation is most horrible 
python. I've patched it just enough to satisfy my needs, so I can make 
nice codeswarm videos, but upstream went dormant around 2012. no chance 
of having that rewritten into py3.

Young startups like Google struggle to find the manpower to migrate, see 
for example Chromium still requiring python2 because ... err... yes.

Most of the stuff from the Apache Foundation is similarly unmaintained, 
but the useful parts are useful so people will use them. Rewriting lots 
of enterprisey code is no fun, so it won't happen soon.

If I'm not mistaken 'offlineimap' is similarly bitrotting away.


At work I have one (1) legacy bit that's used in monitoring that uses 
python2, and it'll age out at some point and doesn't /need/ rewriting 
because, well, why rewrite what goes into the trash can.
(Meanwhile I don't remember any code changes since perl 5.18 or so, that 
stuff Just Works, even Go is easy to update with literally 4 lines of 
code changing from 1.11 to 1.12 - so low-maintenance things exist, they 
are just old and boring and no one talks about them)

Your hallucination that we don't update things is fascinating, but not 
based on any observations.

But there's one difficulty in upgrading: Some migrations are 
destructively expensive, and can't be automated easily. It took us crazy 
long to migrate everything to gcc5 because upstream was unwilling to 
understand dynamic linking. Incidents like that make one more 
conservative ... and it would be nice if gentoo could be at least 
neutral and not act as a damage amplifier for such situations.

Similarly, EOL'ing python 3.5 in Gentoo ahead of upstream caused a 
little bit of friction and forced some people to spend time they had 
budgeted about 2-3 months later roughly now. Which doesn't increase 
their motivation, but who needs friends when you have a nice up to date 
Gentoo.

Would be nice if we could minimize churn, and have reasonable upgrade 
paths. Then maybe our users would spend less time on upgrades and more 
on contributions (or do we want to make things artificially difficult so 
we can pretend to be elite?)


Have fun,

Patrick