Re: SCO: "thread creation is about a thousand times faster than on

From: Linus Torvalds (
Date: Sun Aug 27 2000 - 22:27:59 EST

On Sun, 27 Aug 2000 wrote:
> The beauty of POSIX 1003.13 for the RTLinux side is that we get to
> define POSIX_SINGLE_PROCESS and forbid forbid fork and exec: making
> threads semantics much cleaner.

Well, what you think of as a "beauty" I just consider a silly cop-out by
the standard. Basically, a lot of things can call themselves "compliant to
the letter of the law" wrt POSIX, while still leaving the _user_ out in
the cold.

(This is not something new to POSIX .13, of course. It's just another way
of saying "we've documented what people _do_, and not necessarily given
people a standard way of actually doing things". Which I cannot argue
with: it is, after all, how I think standards _should_ be written. I can
only grumble about it when somebody then confuses "standard" and "design",
and thinks that the two have anything in common).

What POSIX did right was to document existing implementation, and making
people work at making things at least try to make it possible for the
perfect programmer to take differences of implementation into account.
Good and worthwhile goal.

What some people THINK that POSIX did is to design a sensible structure
that everybody does and should adhere to. Which wasn't even the _charter_
of POSIX in the first place, of course. And which would never have worked
out anyway (OSF tried it. May it rest in peace/rot in hell, depending on
your viewpoint).

What _I_ want when it comes to threading is to have the "sensible
structure" in place. I then, as a second-order interest, want to make it
as straightforward and clean as possible (note the "as possible") to
implement a pthreads layer on top of that, despite the fact that I don't
particularly like the pthreads model (*).


(*) At this point, a big "Oh, REALLY?" is heard across the land. Duh.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

This archive was generated by hypermail 2b29 : Thu Aug 31 2000 - 21:00:20 EST