Hurd Traffic #60 For 27 Sep 2000

By Zack Brown

Do you pine for the nice days of minix-1.1, when men were men and wrote their own device drivers?
Are you without a nice project and just dying to cut your teeth on an OS you can try to modify for your needs?
Are you finding it frustrating when everything works on minix? No more all-nighters to get a nifty program working?
Then this post might be just for you :-)
-- Linus Torvalds, 1991

Table Of Contents

Mailing List Stats For This Week

We looked at 341 posts in 1239K.

There were 72 different contributors. 38 posted more than once. 16 posted last week too.

The top posters of the week were:


1. Booting Hurd From CD
5 Sep 2000 - 14 Sep 2000 (3 posts) Subject: "Hurd bootable from cd?"
People: Neal H WalfieldMarcus Brinkmann

Ottavio Campana asked if it were possible to create and boot from a Hurd CD so folks could test it out without actually going through an install. Neal H Walfield replied, "There is one main problem with this: the iso9660 filesystem is broken (at the moment) and does not support symbolic links." Marcus Brinkmann replied:

I looked into it, it is pretty hopeless. We can't use file_start as inode number, because links don't have one, but we can't use easily the directory record offset either, because of the redirected directory feature of RockRidge. But maybe it is possible to work around it, I just don't have the nerves to look into it and grok the rockridge stuff well enough.

Inodeless filesystems are a real PITA.


2. IPv6 For The Hurd
7 Sep 2000 - 13 Sep 2000 (19 posts) Subject: "IPv6?"
People: Roland McGrathIgor KhavkineMark Kettenis

Someone asked if IPVs had been completed for the Hurd, and if it were in progress; and Roland McGrath replied, "Nope and nope. It should be straightforward to add it, however. The pfinet server is written using the Linux 2.2.x ipv4 stack, and it would not be hard to incorporate the Linux ipv6 stack as well. Some glue would be required to have the server handle both PF_INET and PF_INET6." Igor Khavkine felt that the two protocols should probably be implemented in seperate servers. He explained, "It is my understanding that sockets belonging to different families are created with the help of translators hooked on /servers/sock/PF_FAMILY where PF_FAMILY is the number representing a particular socket domain like INET, LOCAL, IPX, INET6, etc. So shouldn't there be a separate pfinet6 server?" Roland disagreed, and said, "ipv6 is a special case because ipv4 and ipv6 are intimately related. I think you want the code to be shared so the server can dtrt for "v4 compat" IPv6 addresses." At one point Igor asked, "in that case would it be possible to hook the same server (pfinet) on two different nodes /servers/sock/inet and /servers/sock/inet6 ? This would be desired so there wouldn't be two instances of pfinet running at the same time. Is this possibility already built into the translator concept?" Mark Kettenis explained, "Yes, this is basically what /hurd/term already does when providing pty pairs. Basically, what you do is setting the active translator on /servers/sock/inet when starting the passive translator on /servers/sock/inet6 and vice versa. There is a possible race when both passive translators are activated simultaniously and I'm not 100% sure the mechanism used by /hurd/term protects against that."


3. Status Of Hurd CVS Tree
12 Sep 2000 (5 posts) Subject: "Some questions..."
People: Marcus Brinkmann

In the course of discussion, someone asked who actually had CVS write access to the Hurd sources, and Marcus Brinkmann replied, "CVS write access have Thomas, Roland, Mark, Yoshinori, Gordon and me. However, Thomas and Roland decide what goes in and what doesn't." Marcus also added that only bugfixes were allowed into the current sources, though he qualified, "Note that a "bug" can be a lot of things, and small, but very useful features which obviously don't break stuff can probably go in as well, but we try to not break things."


4. 'apt' Under The Hurd; Download Directory Reorganization
12 Sep 2000 - 20 Sep 2000 (19 posts) Subject: "apt-get works with a hack (msync)"
Topics: Apt
People: Marcus BrinkmannOliver KowalkeNeal H WalfieldMichele Dalla SilvestraBrent Fulgham

Neal H Walfield posted a two-line patch to get 'apt' to compile cleanly. Long term, he said that either fixing or removing 'msync' support would allow 'apt' to compile cleanly. Marcus Brinkmann laughed bitterly, "Ha! You think if this would be all that is needed, I wouldn't have already committed a change to Jason and get it included? :)" He said that Neal's 'apt' was definitely broken, though he'd look into trying to hack on it further. They tackled it together, and at one point Neal gave a pointer to his version ( . Later on, Marcus gave a pointer to a Debian bug report ( and said, "scroll to the very end of the page (my last two emails). If all three fixes (the above, MS_SYNC, and a shared lib issue) are resolved, apt is ready for the Hurd."

Elsewhere under the Subject: experimental apt packages available () , Marcus gave a pointer to the download directory ( and invited folks to download '*apt*' and install them.

There was no reply to this, but elsewhere under the Subject: hacked apt -> wrong dependencies ? () , Oliver Kowalke reported, "I installed the hacked apt from I set (unstable) in -etc-apt-sources.list. If I do a update/upgrade bsdutils depends on sysvinit which is not available. Maybe is the wrong source?" Neal H Walfield had the painful suggestion, "you need to get sysvinit from Note however, that sysvinit crashes hard for me. THerefore, I recommend that you install to satisfy the dependency and then change the link in /etc/alternatives/tunsystem to point back to /libexec/runsystem.gnu" But Marcus replied simply, "No. It is better to install the bsdutils package from alpha, which doesn't have the dependency on sysvinit (/gnu/hurd/debian). The sysvinit is indeed broken and should not be installed. I removed it now."

Elsewhere, under the Subject: new apt () , Marcus reported:

I changed the setup of, following a suggestion by Neal H Walfield.

The directory structure now resembles a standard Debian setup. You can find the binary packages in dists/unstable/main/binary-hurd-i386/, the corresponding sources in dists/unstable/main/source/.

This directory contains also the changes file, to indidcate which packages are available.

This area is usable by dselect and apt. Use the following apt configuration line in /etc/apt/sources.list:

deb unstable main

For dselect, use a similar setting (depends on the access method).

This should make it more convenient for you to use this repository. Note that some things can go wrong regarding version numbers and dependencies.

You can use several apt lines in sources.list. From sources.list(5):

It is important to list sources in order of preference, with the most preferred source listed first. Typically this will result in sorting by speed from fastest to slow? est (CD-ROM followed by hosts on a local network, followed by distant Internet hosts, for example).

But I think it will still pick the highest version number first, regardless of order in sources.list.

I did not test the new setup, so let me know if apt can't find the files.

Elsewhere, under the Subject: new tar file, please test! () , Marcus announced a new tarball, and Michele Dalla Silvestra reported that 'apt' was missing from it. Neal explained that 'apt' was not included by default, and Michele replied, "Debian GNU/Hurd without apt is not very Debian... :-) Since 15 months Debian GNU/Linux include apt by default..." Neal replied, "Well, apt on the hurd has been working for a week." Marcus was a bit harsher, saying to Michele, "Bullshit." He went on:

if everyone would have helped to port it and fix the remaining problems, we would have had it earlier, too.

In fact, it is due to Brent Fulghams and Rolands effort that it was so easy to get it working now. But the fact is that it doesn't build from the current official source last time I checked, and it requires an libc and two apt fixes to get it working, so I can't upload it to

The tar file will only be made from packages, so it will inherit all bgs in the repository. This will help to not make them forgotten.

There was no reply.


5. PPP: Will The Saga Continue?
13 Sep 2000 - 17 Sep 2000 (10 posts) Subject: "State of ppp"
People: Daniel E. BaumannRamakrishnan MThomas BushnellJames Franklin

Ramakrishnan M and Daniel E. Baumann asked about the status of PPP under the Hurd. Daniel said, "The quarter (at school) has now started for me and this is one of the things that I wanted to work on for my independent study. I am still familarizing myself with the HURD and Mach, but I would llike to know where things stand and possibly who is working on ppp. I suppose the PPP Spec document would also be a good guide in this area?" James Franklin gave a pointer to an earlier discussion ( and suggested contacting the folks involved. He also gave a link to KC Debian Hurd Issue #57, Section #2 (dh20000816_57.html#2) , and Ramakrishnan replied:

After going through the messages some obvious questions are popping up in my mind.

  1. There was a reference to a patch (by Thomas Bushnell) to pfinet to counter its current limitation of handling only one interface (i.e ethernet + loopback). Is this patch integrated into the main source code? If not, where is the patch? Why was it not integrated into the mainstream?
  2. I found that a hell lot of discussions on "HOwto do it" has happened in this list, without any logical conclusion. I do not claim to understand each and every method suggested. Why is it so difficult to fix up a way to implement it?

I think once we fix the way to go, rest of the job will be smooth.

Without PPP, I don't think, individuals (particularly in Asia), will be able to try out GNU/Hurd, and hence the development and usage will suffer. Without PPP, GNU/Hurd will remain a research OS for ever. PPP is an imeediate necessity and we need to act fast to make it happen.


6. Hurd Info File
13 Sep 2000 (2 posts) Subject: "hurd info file"
People: Gordon Matzigkeit

Julio Merino noticed that the Hurd 'info' file was very much incomplete, and offered to help. He asked if anyone else currently maintained it, and Gordon Matzigkeit replied:

I was the last person to touch it. I would work on it more, but got obsessed with how annoying it is to copy information from header files, then have those files be updated and lose synch with the texi.

I experimented with various automatic interface documentation generators, but none were even close to sufficient (generated documentation that was easy to read, and could be integrated in a larger Texinfo work).

I'll try to bring it back into my list of priorities.

There was no reply.


7. Installation Guides
13 Sep 2000 - 17 Sep 2000 (22 posts) Subject: "A new install guide"
People: Neal H WalfieldMatthew VernonJim FranklinMarcus Brinkmann

Neal H Walfield gave a pointer to some work ( and announced:

I have developed a new install guide as the easy guide has become quite out of date. This one was written in the texinfo format so it is also a bit more portable.

I really want feedback, so do not be shy.

Julio Merino gave a link to his own Hurd installation tutorial ( and Marcus Brinkmann pointed out that the page just had a link to an ad. As of KC Debian Hurd press time, however, the page seems to work OK.

Elsewhere, Jim Franklin liked Neal's document, and gave a correction. Elsewhere, Prabhu Ramachandran said he'd installed his system with Matthew Vernon's Easy Guide To Installing The Hurd On Linux Boxes ( and loved it. He said Neal's guide was also good, and might be better for advnaced users. He gave some suggestions for some good links to add to both documents, which Matthew took.

Elsewhere, Matthew also replied to Neal's initial announcement, saying he'd been working on an overhaul of the 'Easy Guide', and added that migrating to 'texinfo' was part of his plan.


8. 'route' Under The Hurd
16 Sep 2000 - 17 Sep 2000 (3 posts) Subject: "route command under Hurd"
People: Tomasz WegrzanowskiFlorian ZimmermannMarcus Brinkmann

Tomasz Wegrzanowski asked, "What's the Hurd way for setting up gateway to the Internet ? something like `route add default gw'." Florian Zimmermann replied:

you can give the defaut-gateway through pfinet, e.g.:

settrans -fg /servers/socket/2 /hurd/pfinet --interface=eth0 \ address= --gateway= --netmask=

I hope there's also a method to manipulate routes manually but haven't found it yet...

Marcus Brinkmann replied:

There is an interface in pfinet.defs but I don't think it is supported at the moment. There is only one interface anyway, so stopping and restarting is an option.

There exists a patch for multiple interface support and route manipulation. It's quite old, and will most likely not apply cleanly anymore, but I assume it will be an issue for after 0.3 is released. People can already start to think about the interface and possible implementations if they are interested in this.


9. Hurd Wiki
20 Sep 2000 (2 posts) Subject: "[Help-hurd] Hurd Wiki (another one)"
People: Eduardo Nahum OchsMarcus Brinkmann

Eduardo Nahum Ochs announced:

I've set up a Hurd Wiki at <>. The main differences between it and the old Hurd Wiki at <> are:

  • the new one is based on MoinMoin 0.2 (= lots of features).
  • there's another cgi, <>, that fetches a copy of all the pages and all the code, so it should be easy to make backups (= what you write there is not automatically lost).
  • it has links to the scripts I used to install it (= easy to install at home, easy to play with).
  • it has ridiculously little Hurd-related content at this moment (= it's a real sandbox now. No one is expected to write anything brilliant there, and people can even use it for questions too dumb to go to the mailing lists).

Marcus Brinkmann replied, "Instead adding yet another doubled ressource, couldn't you create yourself a sourceforge login and let us know the login name? We will then add you to the group of ddevelopers, and you can maintain the one at sourceforge." There was no reply.


10. Mach Docs
20 Sep 2000 - 21 Sep 2000 (2 posts) Subject: "[Help-hurd] Mach Documentation from OSF"
People: Farid HajjiMarcus Brinkmann

Farid Hajji gave a pointer to some docs ( and announced, "FYI, here are a some interesting documents about Mach. Some of them are known to everyone, others may be new." Marcus Brinkmann replied, "That's a lot of stuff! Any recommendations, which are most relevant for the Hurd (basic Mach features which are in gnumach etc)?" but there was no reply.







We Hope You Enjoy Hurd Traffic

Kernel Traffic is grateful to be developed on a computer donated by Professor Greg Benson and Professor Allan Cruse in the Department of Computer Science at the University of San Francisco. This is the same department that invented FlashMob Computing. Kernel Traffic is hosted by the generous folks at All pages on this site are copyright their original authors, and distributed under the terms of the GNU General Public License, version 2.0.