Re: L4-HURD , POSIX, UNIX

[Jonathan S. Shapiro]

On Tue, 2005-11-01 at 13:35 +0100, Bas Wijnen wrote:
> On Mon, Oct 31, 2005 at 08:02:29PM -0600, William Grim wrote:
> > > If you are asking "What would shap do if *he* were the architect?", then
> > > yes, I would begin from scratch -- but I already have done this.
> > >
> > > If you are asking "What does shap think the Hurd group should do?", then
> > > I would say: figure out what your real objectives for the system are,
> > > and from this decide what approach to system architecture is
> > > appropriate. However: several of the truly innovative features that we
> > > have been considering are the same ones that led me to start from
> > > scratch.
> 
> Starting from scratch would mean to create a custom micro-kernel for the
> project as well.  I think this approach would guarantee that the project never
> gets finished.

Agreed. Let me clarify what I mean by "start from scratch". I do not
mean that all code should be rewritten. I mean that principles should be
established first and code should follow from that. If code already
exists that satisfies the principles, or that can be made to satisfy the
principles, or that can be encapsulated in such a way that the
encapsulation satisfies the principles, then by all means this code
should be used,

> I like persistence as well, mostly because I feel it results in a more robust
> system.  However, if people already see concrete problems with it, I'd like to
> hear them now, not after we decided that we want persistence.

I have a coming note on what persistence does and does not buy you. I
spent most of last night in the emergency room with my son, so it did
not go out when I expected it to. Pediatric ERs are not fun places. My
son, thankfully, will be fine, but I am not so sure about the child in
the next room over, who was in some difficulty.

> This makes sense.  However, there is another point: Not only have you thought
> out your views, you also have not thought out the other views as much.

Thank you. This is absolutely correct and very much worth emphasizing.
The two systems that I feel I know very well are UNIX and the
KeyKOS/EROS line.

The thing that I bring fairly uniquely is an in-depth understanding of
security and robustness issues and how to make strong security usable.
Whether this is important to Hurd must still be decided.

>   We
> haven't thought out any of them.

I think that is unreasonably harsh. Some of the people on this list have
extensive experiences with other OS's, and everyone has thought about
the Hurd in some degree. We should reject useful input from anyone.

> I understood that you offered yourself as
> our mentor (correct me if I misunderstood).

You understood correctly.

> > What I would like to keep in mind as we progress is that we need to show
> > real progress. Perhaps a staged delivery lifecycle could help here?
> 
> And more importantly IMO, to show the developers that we're getting there, so
> they stay motivated.

Yes. There are stages in the Coyotos development. I need to write them
down for my clients anyway, and I will send them here as well when they
are ready (soon).

shap