Re: Could be slightly OT

[Marcus Brinkmann]

On Fri, Feb 08, 2002 at 08:51:51AM +0100, Guy Bormann wrote:
> Just out of interest, I was browsing(*) the HURD website and
> stumbled upon this page
> http://www.gnu.org/software/hurd/hurd-paper.html
> concerning the basic design decisions and extrapolations to
> potential exploitation of its features. The idea of file system
> translators for instance is very similar to Hans Reiser's ideas
> about modern file system(s needs) as explained in
> http://www.namesys.com/whitepaper.html

Well, indeed, when Reiser says that we should access things through the file
system namespace, we do this for ports instead using a seperate name server. 
I think it was Roland who suggested this first for the Hurd (I think I
remember Thomas saying that at least :).

>   I think the HURD could be the natural (test)platform to
> implement such ideas instead of having new layers/file system
> reorganisations crammed in Linux for the future releases of the
> ReiserFS.

It is a natural test platform for it!  Even more so, you can develop such
extensions/replacements for the system as a normal user while the system is
running, and in parallel with other users developing different ideas,
without rebooting, stomping on other peoples feet, or requiring extra
permission.

For example, if the hierarchical lookup of ports doesn't work for you, you
can replace the path name lookup mechanism in glibc with something you think
is better (keywords, whatever, I haven't really read Reisers paper, just
glanced over it), you can do it, compile your own glibc, and link that into
programs.  This isn't very hard to do.  I am a bit surprised that Reiser
didn't mention the Hurd, maybe he is not aware of it.  But he is aware of
Plan9, and I think I heard that from the filesystem closure point of view,
Plan9 and the Hurd have similarities.

>   This could be a new incentive for the HURD project

Our incentive is to develop and use this system, and that is what we do!
It's already there, maybe what we lack is marketing or something.

However, the Hurd itself is not a research project, it is a project which
has a usable-in-real-life operating system as a goal.  That's why we are
providing the POSIX API as the main API for example, although it is entirely
optional.

> if this is
> where (database or info retrieval) computing is heading in the
> future. (It might be the wrong impression, but the HURD project
> seems not to catch on as fast as could be hoped for, whatever
> development activity is quietly going on.)

Well, we can only do what we can and invite others to help.  We can not do
what we can and some more.  We would certainly like to see people use the
Hurd for trying out new ideas how to handle file systems and integrate the
system intelligently, and if the outcome is good and usable code, we might
integrate it back into the Hurd itself.  The Hurd was designed to make such
experiments possible and easy.  It goes out of your way.

You might want to write Reiser a reader's letter (hint ;).
 
Thanks,
Marcus

-- 
`Rhubarb is no Egyptian god.' Debian http://www.debian.org brinkmd@debian.org
Marcus Brinkmann              GNU    http://www.gnu.org    marcus@gnu.org
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de