l4-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: A simple question


From: Jonathan S. Shapiro
Subject: Re: A simple question
Date: Thu, 13 Apr 2006 10:48:50 -0400

I'm not sure that I deserve all of the blame for this! :-)

The objectives for Hurd-NG are being set by Neal and Marcus. From my
perspective, I added three things:

  1. I have described how comparable function worked in KeyKOS/EROS.
     It never hurts to know what somebody else has done, but Hurd-NG
     should only adopt these answers where they work for Hurd-NG.

  2. I have sometimes asked questions that exposed contradictions.
     These would have been discovered in any case. I hope that I
     have sped up the process.

  3. I have pointed out when design assumptions would create
     vulnerabilities. Not everybody agrees about the importance
     of this. Hurd-NG must make its own choices, but I note that
     vulnerabilities are very hard to fix later.

Bas is correct that L4/pistachio was not going to meet the needs of
Hurd-NG. The L4 family of architectures was designed as a message-based
microkernel, but it really isn't designed to support object-based
systems. Hurd is not the first group to discover this. The Mungi project
learned this as well. Mungi was able to build an object system on top of
L4, but only at prohibitive cost.

The logical alternatives appear to be:

  1. Build on L4.sec.
  2. Build on Coyotos
  3. Build a new microkernel.

Neither L4.sec nor Coyotos are running today, so both are gambles. Both
groups have a pretty good track record for delivering. How to choose is
mainly a matter of deciding what meets the needs of Hurd-NG better.

Building a new microkernel is certainly possible, but it's a difficult
thing to do. Given that both Coyotos and L4.sec are coming, I do not
think that this is the best focus of effort for Hurd-NG.

> > The desire to create a completely different system only came with Shapiro's
> > influence, way after microkernel reevaluation was already under way.
> 
> That's not how I remember it.  Shapiro's influence was mostly in showing us
> interesting features which we may want to have, and explain what EROS can and
> can't do.  And of course sketch what he expected Coyotos to become.

I certainly described EROS and Coyotos, but I hope that I did not divert
the project. Marcus and Neal described to me certain things that they
wanted to achieve. I suggested refinements, and I pointed out that
persistence has a major influence on how these goals are best achieved.
The issue of persistence is definitely a new issue, but I think this
could be done on L4.sec as well. As for the rest, I *hope* that what I
provided was acceleration in the design process.


shap





reply via email to

[Prev in Thread] Current Thread [Next in Thread]