[linux-elitists] Freeing parts of Unix?

Eugene.Leitl@lrz.uni-muenchen.de Eugene.Leitl@lrz.uni-muenchen.de
Fri Jun 15 14:58:56 PDT 2001

Aaron Sherman wrote:

> Ok, at this point I disagree with you on the basis that a) the MIT
> software that has been discussed here was open source as was Cygnus'
> (now Red Hat) embedded OS.

Sure, Fiasco/L4 is open source, too. No one is using it. The open source 
community doesn't grok the need for nanokernel architectures. In fact
Herr Torvalds himself is sneering down at microkernels.
> Both of these are often held up as good examples of state-of-the-art,
> so if you want to bash the OSS take on real-time, you'll have to shoot
> down those two examples for me with concrete statements.

Concrete statement: user base. Ok? (Mach is not a microkernel, and QNX
is not open).
> The rest of your reply was very difficult to follow in the context of
> my message (which you quote often, but I'm not sure you read).

I read you just fine. I just didn't necessarily followed a point-to-point
line of argumentation, using your post as template. Also, I was getting
drunk at the time. Which doesn not necessarily invalidate, but does
contribute to somewhat decoherence. What was I sayin'?
> I'm not trying to flame you, here, just point out that your real-time
> embedded systems comments are interesting, but you're responding to a
> comment that I made about general-purpose computing. If you and I had

No one needs multitasking. No one needs hard realtime. No one needs deep
embedded capability. No one needs concurrent message passing, either.
Also, you don't need more than 640 kBytes, ever. And the total world 
requirements for a computer installation base are 10, tops 15.

Sounds familiar? 

I'm not saying the requirements are obvious, at this day and age. 
But, nevertheless, you won't find these as the top priorities of
an open source developer at this day and age. Which is imo a mistake
on the long run, and the whole point of that incoherent rant of mine.
And the equally (I've had a few more beers since) incoherent continuation
of said rant.

> been carrying on a conversation about real-time embedded computing, I
> would have been happy to tailor my comments appropriately.
> But, I will argue quite firmly that my statements hold in the general
> case, and that the "right" solutions to your problems are often not
> appropriate for general-purpose computing.

Then, something is seriously foul in the land of general-purpose computing.
> This is a total non-sequitur. How did you think that that statement
> followed mine?

It doesn't. It's a statement, and a true statement. There are no relevant
open source parallel debuggers. Why on earth? Man, this sucks.
> Incidentally, I've heard variants on this statement for 12 years now,
> and so far, no future is in sight where the average developer will
> stoop to worrying about parallel debugging (though I had to do it with
> gdb in '91, and let me tell you, that was no fun ;)

Of course it's not fun, and we've 12 years overdue for parallel systems.
The reason our hardware is so screwed is partly due to our inability
to handle multiple threads of concurrent control.

> Somehow, I thought that we were having a discussion (not even
> necessarily a debate) about the useful bits of existing UNIX(tm)
> source code. Random, and unexplained ad hominems are severely out of

Ahh, you see, that's the second mistake you made ;P

> place in such a context.
> The statement you replying to contained the following concepts:
> 1. not UNIX-like
> 2. VMS-like
> 3. per-installed-file permissions needed
> 4. granular permissions needed
> Which of these did you disagree with? Or, did you have some more
> abstract problem with my statement (spelling, perhaps)?

Chiefest quibble: insufficient blood alcohol content.

Hic, haec, hoc. Hick.

-- Eugen* Leitl
ICBMTO  : N48 10'07'' E011 33'53'' http://www.lrz.de/~ui22204
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3

More information about the linux-elitists mailing list