How to contribute to GNU Hurd (2006) | Hacker News Reader

How to contribute to GNU Hurd (2006)

gnu.org

58 pointspmoriarty15 comments

Every time I checkout GNU Hurd I'm flabbergasted that it is still 32bit. It is something that is really hard to get past and honestly I don't have the skills to contribute to make it 64bit. But I just don't have time for something that is limited to 4GB address space[1]. Too far away from modern to be honest.

(IMHO)

[1] https://www.gnu.org/software/hurd/faq/64-bit.html

Jesus, are you serious?

Even tablets are 64bits nowadays...

I'm sure if we donated enough it would be ported in a month.

So if you were to use the hurd, you'd have to run n two-gigabyte shards instead of n/4 eight-gigabyte shards. This is not a big deal. The hurd has debilitating problem, this isn't one of them.

So there’s the “how”, where’s the “why”?

Why Hurd and not Linux? Or one of the BSDs? Or any of the thousands of OSS projects actually useful to real people in 2017?

Different paradigm. The translator idea is cute. Samuel demonstrated the benefits of mounting remote large gzipped isos and accessing the data without local fetch (because the translator could just enough to get the index and then allow per file access).

Fine grain virtualisation is pretty hackish in Linux, but it's part of the design of the Hurd.

On the Hurd you can trivially virtualise the filesystem (on Linux you need FUSE for that but it doesn't go far enough), there are user namespaces by default (on Linux they are a hack and that makes local privilege escalation bugs really bad), drivers run in userspace, so different users can have different views on the hardware, etc.

All of this is quite relevant today, especially on shared systems such as HPC clusters.

The Hurd is the only multi-server micro-kernel system with a Unix persona (other micro-kernel systems are either single-server systems or don't bother with a Unix-like persona). I find it a lot more elegant than Linux, which needs to move more and more things into privileged kernel space.

With the NetBSD rumpkernel the Hurd can reuse NetBSD device drivers in userspace.

There are a lot of things to like about the Hurd. If only it had more contributors! It's a great project that provides practical software freedom (like other high profile GNU packages), thereby liberating users from the root user.

Rather than trying to go mainstream or catch up to Linux, they could instead serve as a testing ground for experimental OS algorithms.

They are notoriously hostile to first time contributors, don't hold your breath.

From which decade is your experience?

This document explicitly states that they encourage new contributions, there is a list of "easy" issues to work on, and they apparently participate in Google Summer of Code.

Of course reality might look different, but I'm willing to give them the benefit of the doubt that they mean what they write.

Could you elaborate please? What are your experiences?

(2016)

Given this is about the Hurd, there's a non-zero chance it could have been (1996), so I'm fine with a 2016 article...

According to the bottom of the page (as well as https://web.archive.org/web/20081217030432/https://www.gnu.o...), it goes back to 2006. It could be older, but I suppose we can put 2006 in the title.

Nobody would even remember Hurd if it wasn't a GNU project. TempleOS has more organic interest than Hurd. Urbit has more organic interest than Hurd.

Reply on news.ycombinator.com