Re: Confusion about where to go in Hurd

[<olafBuddenhagen@gmx.net>]

Hi,

On Thu, Jul 26, 2007 at 10:19:01AM -0700, Andrew Sackville-West wrote:

> There are a series of branches of GH that are independently pursuing
> the same goal: a completely GNU OS. There exists lots of crosstalk
> between those branches of GH such that none of them are particularly
> ahead or behind the others, just different with different ideas of the
> path towards the same goal. Possibly even, they have different ideas
> of what that goal is though nominally it is the same. 
> 
> Some of these branches continue to pursue work in directions that are
> largely considered to be dead-ends, though the work completed in those
> directions is expected to be at least informative of other future work
> in other directions, if not downright portable to that future work.
> Other branches are pursuing what are considered potentially more
> successful directions of the development and at this time look to be
> the future of GH. 
> 
> But as has been shown in the past, the direction taken may not end in
> success (which iteration of new (micro)kernel are we at now? 3 or 4?
> trix/BSD 4.4-lite -> mach -> l4 -> other?) so the other, less likely
> branches continue as they are already functional (for some definitions
> of functional) and may in the future continue to be useful. 
> 
> is that a reasonable bird's-eye view? if not, kindly please correct me

Yes, it is a surprisingly on-the-spot description of the situation I'd
say; except for one point: In your remarks, you seem to imply that there
are actually several actively developed Hurd variants. This is not the
case.

There is the mainline implementation using gnumach, with slow but
ongoing development. There is also X15-Mach, but this is not any new
design -- it's supposed to be a better implemented drop-in replacement
for gnumach.

The original Hurd/L4 effort has been abandonned long ago, because of
some fundamental problems with the design. ngHurd, which was supposed to
be the successor, never came up with any code or even a fixed design;
and even the design discussions pretty much died away after Marcus
realized that the strong focus on Coyotos might not have been the right
way to go after all...

(While this dying-down previously has been attributed to a lack of time
by the main architects of Hurd/L4 and ngHurd, my impression is that
after the repeated past disappointments, Marcus now prefers keeping more
in the background and trying out his new ideas on his own, rather than
setting the direction for others...)

So Hurd/Mach is really the only active variant of the Hurd. If you want
anything else, you need to take it up yourself.

-antrik-