Gentoo Archives: gentoo-dev

From: Gabriel <egabriel@××.com>
To: gentoo-dev@g.o
Subject: [gentoo-dev] Hurd
Date: Thu, 25 Jan 2001 10:47:36
Message-Id: Pine.LNX.4.21.0101251130230.16681-100000@fnord.io.com
In Reply to: Re: [gentoo-dev] Re: Tom by drobbins@gentoo.org
1 > A question -- could you give us a quick overview of the differences between HURD and
2 > the Linux kernel? I'm curious about what new things HURD offers.
3
4 This deserves a longer answer and I will post one when I have a bigger block
5 of time, but a few observations:
6
7 The hardware support is still much worse than linux. There isn't incentive to
8 do too much there for the Hurd team yet (chicken and egg... need a dist before
9 anyone really cares, need drivers to make a dist interesting enough to care
10 about). It supports a modest list of ethernet cards and scsi controllers,
11 and with X can support whatever video cards X can, but it as of now has no
12 sound driver at all.
13
14 Things are funky because it is a microkernel. For instance, there is a login server, which runs under the 'login' id. You login at a prompt like
15
16 login> login myusername
17
18 This is different than the unix 'getty' stuff.
19
20 A great deal of commands you normally run (mount, ifconfig, etc...) are obviated by a new command, "settrans". This is because ext2fs, cd9660, pf_inet, etc, i
21 are translator daemons, not just devices in the unix sense. This "herd of
22 daemons" is what really makes this different from Linux. The kernel is
23 very minimal. These daemons do sort of what modules do under linux, though
24 the analogy isn't really appropriate. One of the side effects of this
25 is that the daemons are proper programs. You can run them with --help
26 apart from settrans, and they will list their own options. How often I
27 wish I could do that with a module.
28
29 Another good thing about this is people can develop kernel version independent
30 modules; making the distribution of drivers a lot simpler. Ironically, this
31 will probably encourage the distribution of binary drivers if Hurd does
32 catch on.
33
34 There are some things that don't work as of now. Irritatingly, one of them is
35 df. Booting from CD is also a problem. Both of these apparently stem from
36 the way inodes are currently handled. These aren't insurmountable, but it
37 doesn't look like a lot of process has been made. It is currently x86 only
38 as far as I know. The limited hardware support is reminiscent of the
39 oldest versions of Linux.
40
41 BTW, I am not a Hurd developer; just a guy that installed it for kicks, so if
42 my information is not factual, please keep me honest. I am enthusiastic about
43 this because the most recent debian builds are better than I think a lot of
44 people realize, and there is a lot of opportunity to shape the way it evolves,
45 and the Portage system sounds like the right way, the GNU way, to do it.
46
47 --Gabriel

Replies

Subject Author
Re: [gentoo-dev] Hurd drobbins@g.o