Re: Future Direction of GNU Hurd?

[Paul Boddie]

On Friday, 11 September 2020 18:47:55 CEST Marc Dunivan wrote:
> To Whom It May Concern:
> 
>   I understand that the I4 micro-kernel for GNU Hurd is dead, and that
> micro kernels beyond GNU Mach are just research.

Well, it would seem that there is no apparent interest in L4-family 
microkernels for the Hurd, at least from its remaining developers, and so 
other microkernels have only moved beyond research and into actual products 
outside the realms of the Hurd.

Although one might think that there is some kind of community around the Hurd 
that might entertain discussion and experimentation around operating system 
design, a perusal of the mailing list archives mostly reveals plenty of 
brainstorming, speculation, the occasional blue sky project being spun off and 
losing momentum, plus a degree of infighting, interpersonal conflict and 
impatience towards people not mucking in to get the Hurd "finished".

While all of that was going on, a few of the L4 microkernels actually became 
available. Sadly, public discussion around those is pretty muted even amongst 
the more prominent ones. My guess is that people are busy using them in niche 
applications.

> What is the GNU mailing list to follow for the future direction of the GNU
> Hurd Project?  ( What micro-kernel?  ARM AARCH64? USB Hurd server support?
> Graphics processor support? )  It is unclear to me as to which documentation
> one should be reading and what physical hardware is being targeted as
> reference platforms for developers to use, and for what purpose?  What are
> people working on?

You might want to follow the bug-hurd list on the same mailing list server for 
Hurd development discussions. The list name is misleading - you would think 
that it is about bug reports - but I guess it started out as some kind of 
joke.

> Is its use for a graphical-shell personal computer still even being
> considered?

My impression is that you can use the Hurd today for a graphical desktop 
experience, but there are a number of limitations around the kind of hardware 
that will actually support it. A few challenges that I can think of are...

Portability: Mach is supposed to be portable, but I doubt it is getting the 
exercise it needs any more. Also, multi-core/processor support is only in its 
earliest stages in the Hurd, as I understand it. I imagine that various Mach-
based systems did support SMP in that I can't imagine things like OSF/1 (Tru64 
Unix) not supporting it, but there is likely to be quite a bit of work needed 
to recover (or introduce) support for such lost features.

Hardware support: I don't know whether USB devices are supported, for example, 
but a while back there was limited or no support for them in the Hurd. The 
common thing many projects of this nature do is to try and pull in driver 
support from other places (typically Linux or the NetBSD rump kernel stuff), 
but I remain fairly unconvinced that this is the magic bullet everyone thinks 
it is.

Documentation: as you may have noticed, documentation for the Hurd appears to 
be some very old white papers, some patchy reference documentation, and a 
bunch of IRC transcripts on a wiki site.

Social factors: you will see plenty of inertia or what people used to call 
"stop energy" in the mailing lists. Just review the previous messages on this 
list or look at recent messages on bug-hurd for examples. One gets the 
impression that unless someone is either a top-level operating system 
architect and/or willing to burn themselves out fixing up huge amounts of 
existing code or writing new code in precisely the way envisaged by others who 
nevertheless aren't going to be writing any themselves, then new contributions 
are not really welcome.

I recently read an article about Linux kernel development where someone 
actually wrote that if something or other made kernel development easier or 
more accessible then it would be attracting the "wrong kind" of people. Well, 
that would seem to be the view of some people in the Hurd development 
community, too, judging from recently expressed sentiments.

Of course, many other people in the wider world realise that successful 
projects require many different types of contributors in order to actually be 
successful, and judging by the increasing stagnation of what was a fairly 
well-resourced and well-publicised project, I think we can guess whose 
viewpoint is the more realistic of the two.

Paul

P.S. These days I just do my own thing with the Fiasco microkernel and L4 
Runtime Environment. Maybe it will become something more substantial one day, 
but in pursuing such solitary activities, at least I have no expectation of 
there being some kind of forum of similarly-minded individuals when, at least 
in the context of the Hurd (and, for the most part, Free Software generally), 
no such forum would seem to exist.