Re: Changing from L4 to something else...

[Jonathan S. Shapiro]

Richard:

Regrettably, Alfred has asked you a question out of context and without
regard to fact. Nobody has proposed switching Hurd to EROS. As "the EROS
guy" I would be inclined to advise strongly *against* such a plan.

Here is a more complete sense of what is going on over in l4-hurd land.
My suggestion is to let the group sort it out, and try not to let
yourself be trapped by Alfred (or anyone else) into issuing opinions
based on partial and misleading information.

Hope all is well with you. It has been rather a long time since we have
spoken.

Regards,


Jonathan Shapiro



>From the perspective of most FLOSS users out there, the original
*technical* goals of the Hurd (a free, extensible POSIX system) have
been met by Linux. Not precisely in the way that Thomas or RMS
envisioned, but well enough for the needs of much of the world.

This has left Hurd struggling for a compelling mission. Progress on the
Hurd has not been held up by switching kernels. It has been held up by
the absence of a clearly articulated, consistent set of architectural
goals that would make Hurd useful, interesting, and compelling.

In my opinion, the Hurd project is coming to a cusp, and it faces a
choice between two options:

 1. Hurd can continue to pursue the goals for the GNU OS that were set
    out in 1983. This is a fine thing to do, but in my opinion this is
    basically an obsolete activity.

 2. Hurd can try to establish a forward looking vision that makes sense
    from where we stand now in 2005.

I do not think that there is a consensus among the l4-hurd list
participants yet about which way to go. It appears to me that people are
split about evenly, with some preferring to enhance and continue
"current hurd" and others wanting to look forward. Both positions seem
to me to have merit.

The issue at hand isn't really about "switching kernels" at all. In
reality, using L4 for Hurd was never a practically feasible option. It
was simply bad timing. The L4 community made an incompatible transition
in their architecture just as the question came up, and the timing of
their decision came at just the wrong moment for l4-hurd. The EROS
project made a similar break at exactly the same time (in fact, the EROS
and L4 groups collaborated on the respective revisions).

The realistic choices today would be L4.sec (the L4 successor) or
Coyotos (the EROS successor). Both remain works in progress, but both
will complete fairly soon, and at least one of these groups has
demonstrated that it is willing to collaborate effectively and usefully
with the Hurd team.


You replied to Alfred:

> I am against it--to keep changing platforms means never getting it done.

I agree. But the reality is that no feasible platform has been selected,
and there is a bigger issue to consider: until Hurd figures out where it
is going, it has absolutely no chance of getting there. Hurd must first
determine the objectives and then choose the appropriate tools for the
job.

Over the last month, a discussion has occurred on l4-hurd that has
reality-checked some of the originally stated Hurd goals and gone
through various elements of the current Hurd design to show that
suitably updated goals simply cannot be met with the current design.

This discussion seems to be wrapping up, and Marcus and Neal are now
trying to coalesce it into a coherent statement of goals and direction.
At that point, the Hurd community will need to decide what path to take.

My suggestion is: hold off for a bit and see what emerges.

Best regards,


Jonathan