Web lists-archives.com

Re: Hurd-i386 and kfreebsd-{i386,amd64} removal




Carsten Schoenert, le sam. 13 avril 2019 12:41:25 +0200, a ecrit:
> Am 13.04.19 um 12:06 schrieb Samuel Thibault:
> >> Both architectures haven't seen any major development in the past years
> > 
> > They have.
> 
> O.k. need to be more specific, so the same as you mentioned further
> down, ..."in the context of Debian, the packages of Debian and it releases".

In the context of Debian as a distribution itself, there is not much
more to be done actually: the distro can be installed with the normal
debian installer, we don't depend on unreleased patches to have a
working system. We even have gotten llvm working recently.

In a broader Debian meaning, for instance with Helmut we have achieved
cross-bootstrappability of hurd-i386 from amd64, which is really a great
thing, because we know that we can now reboostrap the whole distribution
thanks to this if needed.

> > Patching software should be handled upstream indeed.
> 
> Yes, but most upstreams are a bit reserved

You mean, more than the Debian maintainers?

Well, at some point, "so be it, you won't have the works-on-Hurd badge".

For base packages like librsvg, the question is of a bigger importance,
for the port of course, but also for computer science in general: if
base packages can't easily be ported to new operating systems, the whole
computer science will be just stuck with Linux, and I don't think it's a
good thing.

That reminds me a recent paper about the requirement for fork() in the
Unix interface (https://lwn.net/Articles/785430/), which notably says
that because of the complexity for implementing it, it's hard to create
new operating systems with new ideas while providing a POSIX interface
for being useful in general.

> >> So I disagree on "One person is enough"
> > 
> > I meant only for the Debian-specific things, I am the only DD who
> > currently uses its key for signing packages, making CD images, etc.
> > That's what I meant by "the daily ports things".
> 
> Well, I guess it's not that easy I fear as there are no parts that can
> be seen as separate standalone things, it's all connected in various ways.

Yes, these are very intertwinned, but I like working on it and the
current Debian infrastructure makes it easy enough to do.

> But realistically it's not enough in my eyes to keep Hurd on even
> tracking the normal evolving of Debian.

I have since long stopped hoping that the Hurd port would ever be an
arch released in Debian (see previous threads in the past years about
moving to debian-ports). Just for the security guarantees it would
require, that can't work.

But as a debian-ports, I believe it can continue working just like it
has in the past years.

Samuel