Re: The idea of an own L4

[Jonathan S. Shapiro]

On Sun, 2005-10-09 at 16:46 +0200, ness wrote:
> This needs some clarification. Some people understand that as the 
> dresden/karlsruhe guys are unfair/don't want to help us. But there are 
> basically two problems. On the one hand side, they're working on a 
> university. So sometimes they _can't_ help us (e.g. give us dev source 
> as one is writing his/her thesis).

This is absurd. There is absolutely no reason why active research should
preclude publication. Indeed, the entire point of public research is
public results. As a counterexample: every single commit that has been
done in the EROS and Coyotos projects since 1991 has been public since
the instant of commit.

There is an qualifier to this statement: occasionally we have said "let
us try something that may not work." These activities are usually done
on branches to prevent breaking the main path. There have been times
when we have forgotten to make these branches publicly accessible. Most
of them died within 48 hours, so it didn't really matter.

But I want to emphasize the word *forgot*. All of our effort, as a
matter of policy, has been executed openly.

>  On the one hand side, we have 
> different goals. We want a kernel that matches our needs in a pretty 
> way.

Respectfully, no, you do not. You want a kernel that does the right job
in the right way with an interface that meets your needs. This is quite
different from "pretty", as our discussion of capabilities is
demonstrating.

The one place where research may be getting in the way is time: a
research group deliberately sets aside market pressures, and takes the
time that is needed to get the job done right. This may or may not be
fast enough to meet your needs, which is a legitemate objection.

But if you want to fix it, the place to start is to determine what your
realistic needs are. What is the Hurd schedule, and when do you need a
kernel? Perhaps they can meet it, and perhaps they cannot. If the
schedule is purely arbitrary, do not expect your suppliers to pay any
attention to it.

It *is* true that the Dresden and Karlsruhe teams have been very bad
about opening up their discussions -- we cannot even convince them to
use their own mailing lists for discussions so that we can participate.

I would add: the EROS/Coyotos efforts made an explicit commitment to
"developing naked and in public." This is a very hard move to make for
many reasons. I believe that Hermann would do this if he could, but the
German research funding model does not make it easy for Dresden to do
this.

shap