Christopher Allan Webber

GNU Hurd, Mach, MIG release new versions

Christopher Allan Webber at

New versions released of GNU Hurd, GNU Mach, GNU MIG!

There's a lot of passive hate for the Hurd out there, but I think this is mostly from people either criticizing the time it's taken to deliver (which is fair, though you can run it... just not as much as you'd like :)), but weirdly there are also people who seem to be criticizing the architecture, which I don't understand really. IMO, the Hurd design is great, and if we were all able to run it today, we'd probably be in a lot less trouble than we are today. There would be no need for kdbus since the Hurd's design centers around safe and asynchronous IPC, we'd have a capability security based model would increase safety and make a lot of the "oh no how do I isolate things" concerns greatly reduced, and other things. The systemd debate wouldn't be happening, because componentized daemons for many of these systems would already be baked into the OS design.

In an alternate universe, these things are true. In the meanwhile, GNU Hurd is still to this day an adventurous design. I wish it the best!

Tsyesika@io.theperplexingpariah.co.uk, mnd, mray INACTIVE, Douglas Perkins and 8 others likes this.

Claes Wallin (韋嘉誠) shared this.

Show all 7 replies
AFAIR Marcus Brinkmann and Neal Walfield had strong criticism on the Hurd that is was inherently hard to support unix compatility security. Has this been addressed meanwhile? http://walfield.org/papers/200707-walfield-critique-of-the-GNU-Hurd.pdf

Bernhard E. Reiter at 2015-04-16T15:49:06Z

Christopher Allan Webber likes this.

Yeah, that's known as The Critique. I will admit I don't know much about it, except that some present Hurd developers it is capable of being overcome, given someone puts enough time into it.

The other main criticism seems to be from a microkernel perspective, but we're more or less implementing something that emulates a microkernel plus daemons on top of a monolithic kernel by adding systemd, d-bus, containers with sandboxing, and the like. The best resource I can find addressing the subject is here, which seems to find that the known problems are solvable but we don't have proof without real implementations, but that's from 1996.

Christopher Allan Webber at 2015-04-16T15:57:32Z

Claes Wallin (韋嘉誠), Charles Stanhope likes this.

It's great that people are continuing to work on GNU Hurd. In the very worst case it is an alternate history retrocomputing curiosity, and that's fun. In the best case I believe there ought be a lot more computational diversity AND a lot less proliferation of projects that are barely different from each other.

Mike Linksvayer at 2015-04-16T16:52:57Z

Elena ``of Valhalla'', FLWNQWUD, Claes Wallin (韋嘉誠), Free Software Foundation and 1 others likes this.

Hurd will always have a soft spot in my heart. :)

Sean Tilley at 2015-04-16T20:50:01Z

mnd, Claes Wallin (韋嘉誠), mray INACTIVE likes this.