Re: Not offtopic: Linux ! Co na to Haynie?

[Bob <NAkEd.SkatEr@BigFoOt.CoM>]

>No ja to beru tak, ze ten samotnej kernel NENI uplne vsechno. Teda spis
doufam :-).
>jedina sance jak udrzet a hlavne rozsirit user a developer base je vyuzit
momentu,
>kterej ted ma linux. Aby bylo jasno, tak bych taky radsi videl QNX, ale
zaroven
>chapu i tenhle uhel pohledu a naprosto uprimne si myslim, ze to muze dat Amce
>novou sanci. Vem si to i tak, Be je dost dobrej system, ale skomira na
nedostatek
>developeru (a podle me i nedostatkem driveru). Tohle Linux muze poskytnout.

Z Linuxu nikdo neni moc nadsenej (mimo Collase). Tady jsou vystrizky majlu
od Dave Haynieho.          
---------------------------

You clearly don't understand the role of the kernel in an OS. The Linux
kernel today may meet a modern standard of size-wise efficiency, but
lacking realtime, it doesn't meet any modern standard of time-wise
efficiency, and lacking threads, it doesn't meet any modern standard of
CPU-wise efficiency. 
------------
Well, you know, I always favored a next-generation Exec than either QNX
or Linux. But given the choice, QNX is quite Amiga like in the way it
works, and Liunx is the opposite of Exec. This is a classic no-brainer.
---------
Linux is uninteresting not because Linux is junk (it isn't), but because
it serves the needs of multimedia and desktop work less well than today's
AmigaOS. And most of use were hoping for something far better, if not
absolutely revolutionary, for the AmigaOS of the future.

Maybe a Linux-based AmigaOS would make a better server than Windows NT. Though
in the recent tests, NT won; my guess is that this is based on lots of
multithreading in NT versus traditional expensive processes or polling in
Linux, typical UNIX-style coding, but I'll admit to not pouring over the
results. The Linux folks are. I have no doubts that the Linux community
can outperfom Windows if they put a mind to it. I can outrun the old guy
who owns the tree farm South of mine if I had a mind to do it; the
question is, would anyone call that a real accomplishment. Very primitive
BeOS web tools were running twice as fast as their MacOS versions with a
fairly simple port done (no Windows tools, at the time, had been ported).
That's a good indication that you have something.
---------
In Linux speak, user-level threads are NOT threads, in any way, shape, or
form. They're nothing more than coroutines managed by a linker library,
no preemption involved. You had these in Liunx 1.x, in MacOS, in CP/M, on
the C64 (built in on most of the later FORTH systems). You could hack up
something like this is 20 minutes (well, I could).

In fact, they do also have what everyone else in the world calls
"threads" and Linux folks call "kernel-threads". These are, of course,
lightweight, pre-emptive, kernel managed threads, rather than application
managed threads (I did some digging). The problem is, since the Linux
kernel isn't reentrant, there are few cases where these actually allow
multitasking; much of the time, a kernel call in one thread will block
all others. Some of the talk I've run into on the Linux message boards
claims that no one uses kernel threads in serious applications. 

And realize, also, that what they call "kernel threads" and everyone else
calls "threads" isn't where this all ends. A modern OS will have
kernel-MODE-threads -- the kernel itself will run any number of threads,
all in kernel mode as opposed to user mode. Linux does not have this (far
as I can tell; one of the tricks in evaluating Liunx is that there's no
place that has any real answers, and yes, I have checked placed like
slashdot.org, etc). 

> I bet you 10 to 1 that there'll be a true blood operating system war between
> QNX and Amiga if/when they are released. 

Maybe, I dunno. BeOS has far better architecture than Linux, at least as
good as QNX at the low-levels and better up top; far, far, better
development environment and API, and more momentum than either a
quasi-Linux AmigaOS or QNX on the desktop today. I'll need a reason not
to move to that 110% pretty soon.
----------
> Heck, actual development of the kernel wouldn't hurt either, but it may be
> too early for that yet.  I'm still working on it, adn it's still a moving
> target.  But, hey -- you never know.  :)

I never gave hope to "letting THEM do it"; not today, not in '83... I
wish I could be wrong about this more often -- I still can't believe that
Amiga and the Commodore/Amiga that followed was all THAT unique, but it's
hard to find something to suggest otherwise. As nice as that might make
some feel about their role in Amiga, it basically sucks. All I ever
wanted was a better personal computer, I didn't much give a damn who made
it happen (yes, it was fun beyond words to be one of those guys, but
that's never been a necessary precondition to the creation of a better
personal computer, at least in theory). 
---------
What you have these is pure commodity software; it has
virtually no serious demands on any part of the system. You could do a
reasonable pass of that kind of stuff on an Apple II or C64 (and it fact,
did). You can't power a 48 track digital audio workstation from a C64.
That's the kind of place you'll see the difference between a Linux kernel
and a QNX kernel. For everyday Mom & Pop stuff, any old 133MHz PClone
would do the trick (he says, as the owner of a 133MHz PClone laptop which
pretty much does the trick for the things given it). 
---------
> > Hello ? Because THEY SCREWED US, that's why ! They promised us QNX
> > many months ago, and now suddenly they switch to Linux because Jim
> > Collas (who apparently is no techie) found it's the hip thing at the
> > mo'.

> But he _is_ a techie. 

But one does wonder. He's 38 -- my age. He went into management 10 years
ago; maybe the fast track to money, but way to early for any techie worth
his salt. We simply don't care all that much about money, doing the tech
stuff pays well enough, and it's not just fun, it's a passion you can't
just switch off at will. Anyway, he seems to have designed DSP systems,
gate arrays, mainly hardware (don't know who for). Also some
software/games for the C64 and other 8-bits. 

However, hardware design and software hacking don't make one capable of
making complex and involved software decisions. I feel I can comment on
software because I've actually done it -- I even have a degree (doubled
up with EE at CMU). Just being a HW guy, you'll do software, but not
necessarily software engineering.

Thus, one must conclude [a] Collas is a really bad manager, an empire
builder, and making bad unilateral decisions just because they're his, or
[b] he's taking full advice from his technical staff, and weighing that
against the advice from his marketroids. I'd bet on [b].

So then the problem comes down to who's advising him, technically, and
how much does he lend that against the marketing advice. I know if I
came in, as the tech guy, and said "what marketing wants just plain won't
function", I would have the boss's ear. I similarly expect that if the
marketroids came in and said "we can't sell more than six copies of this
new OS", they'd have the boss's ear. 

So I don't know. Could be Linux was dictated as a marketing requirement
-- I have know horrible managers, such as Bill Sydnes, who have done
things like that (the IDE and the casework were both dictated to use by
Sydnes for the A4000, no way out). I would like to think they're rare and
that Collas is not doing this. 

I know they have some ex-Apple people in there. Some people at Apple
thing UNIX is a perfectly reasonable thing for a multimedia machine (to
the tune of US$450M, when they bought NeXT). I hope none of them are
techies, but then again, if all their life's been lived under the MacOS,
maybe Linux actually is better, relatively, though I can't really
believe that. Another theory is that, based on the rampant attrition and
repeated firings at Apple over the recent years, they may just have too
many bozos on their bus. But that's today -- could be they have a good
guy in there, from the early days when Apple actually had something. 

Then there's Dr. Allan. I really don't know that much about his plans and
ideas, or how he actually did running Amiga Software those couple of
years. He did have some good ideas for the Hombre-as-game-machine, but
when you're not actually shipping a real OS, you're solving an entirely
different problem. To some real extent, the HW and SW groups actually ran
themselves -- each module owner was responsible for the module, not
management. So I really don't know, but my impression has always been "A
Good Guy". 
-------
> > QNX is a really cool idea (based on technology that began evolution
> > in the 80's and is still evolving), that fits into several really
> > cool niches 

> As someone that has used Linux on both the Amiga and PC hardware,
> most of the comments about Linux are way off base.  How many have
> used it for projects?  I have. 

Same here. Linux was the original OS for the Met@box. We canned it last
fall, in favor of OS/2 on the high end, QNX for the low end (I rather
expect we could have done the whole thing on Neutrino, had that been
mature in time). 

> How many have measured the performance of Linux versus Window 9x?

Not an issue -- my new cat could write a better OS than Windows 9x, if
only performance were an issue. Yes, Windows sucks. That makes everything
put up against it look good. But I can guarantee you, Linux will pretty
much such doing Amiga like things too, if you chose any metric other than
"does it look ok next to Windows". That might be all you need for a
marketing driven decision, but it's a zillion miles away from pushing the
envelope.

> There is no real advantage of going with QNX over Linux.

Maybe not if all you're writing is a Web server, but if you're doing
multimedia, it's day and night. I'll take the realtime, multithreaded OS
over the non-realtime, non-multithreaded OS any day of the week. If you
don't get this, you don't get Amiga.

> However, right now, Linux is hot.  Everything from Oracle to
> EDA applications is being ported to Linux.  You can get WordPerfect
> for Free with Linux...

While this is all true, what does that have to do with Amiga? If they're
just using the Linux kernel, and building up some grand new NeXTian object
and GUI overtop (of course, with the Amiga name attached), nothing
outside basic shell tools and device drivers will run (and the latter
only if they specifically DON'T go in and fix the re-entrancy and
multithreading problem in Liunx). If they deliver a bog-standard Liunx,
that stuff will run; but it runs today on my Red Hat machine, so why
bother with Amiga? 
 
> Yes, for Amiga Inc. to change horses in mid stride seems stupid,
> but I am willing to wait for more information and not make a knee
> jerk reaction and scream foul. 

The foul is entirely appropriate when you say "Amiga" and "Linux" in the
same breath -- they are as close to opposite philosophies as one is
likely to find in an OS these days. Gateway doing "Gateway Liunx", I have
no problems with that, I wish Linux all the success in the world. Just
don't even pretend you're getting AmigaOS out of a Linux kernel. It will
not happen.

> and while you might feel personally betrayed
> by the choice, stop and think why it is that you feel this way?

I would guess it's the 11+ years I devoted to making the Amiga what it's
known to be, being essentially mutated into something totally not Amiga
overnight, that leads to my sense of personal betrayal. But that's just
me....
-------------
> Linux has public visibility, QNX doesn't

As hype machine food, I agree, it's a move that'll get reported. They'll
build an AD2000 version of the NeXT OS -- standard UNIX underpinnings
with a potentially neat object layer that no one will ever support. This,
of course, won't come out in the early days. 

Their alternative is bog standard Linux, to remain compatible. Nothing
there but perhaps some bug fixes, a GUI tweak and standardization,
basically the same stuff Red Hat and Caldera can do. Nothing wrong with
that, if it were "Gateway Linux" or some such, but it's very, very wrong
to pretend this has anything to do with Amiga. 
-------
>   I don't understand these remarks.  QNX is a Unix, probably much more
> like Linux than the Amiga Exec.

No it's not, not even remotely. Have you even studied QNX? It's very much
like a protected-more Exec. Yes it's using the POSIX API, but that's ONLY
an API. Architecturally, it's everything like Exec and not even remotely
like Linux/UNIX. If you don't get this, you're not going to grok the
discussions that follow...
-------
> On other fronts, I am seeing a "Well, I guess Linux isn't so bad"
> attitude.

Ask 100 random people on the street corner, and you'll get a pretty high
"well, I guess Windows isn't so bad" response. People who don't know why
their system works the way it does won't understand the details of why it
doesn't work when it doesn't. Or why one kernel will produce dramatically
different preformance results than another.

Unless we educate them. So you don't understand, listen. If you still
don't, ask. It's important for everyone in the community to understand
why Liunx is not AmigaOS, and how this will hurt our concept of Amiga. We
experts will help, answer what we can, and wait to tear apart the
forthcoming Amiga [the company] white paper which proports to explain it
all in a way we'll find tasty. I rather think I would be more impressed
with a marketing paper -- at least I can understand selling out the Amiga
for cold, hard cash. I wouldn't like that, but if they honsestly think
they can't make a go of doing it right, they should have the cojones to
say that outright, not hide behind a mask of hoped for technical
befuddlement (well, I guess I'm buring a bridge or two in all my rants,
but sometimes there's nothing like a little pyromania to catch some
attention). 

> I don't care HOW good Linux is at being Linux, it is going to SUCK at
> being AMIGA. 

Amen, brother Greenie!

> Polite, my ass. Both paths possible, my ass. Choose up! Only the highway 
> needs a yellow line running down the center.

And I hope not to see any runnin' down a back or two, at least here on
the list. Fighting for what's right ain't always easy, but you're always
damn glad you did it when the smoke clears. 

It could be a battle. I can imagine the Amiga user groups, weary already,
looking at Linux as a compromise to get them on the map, industry
speaking. When you're dealing with details as low-level as kernel stuff 
(and I'm damn glad to have a few knows-way-more-about-Liunx-than-Haynie
types here on the list pointing out new flaws I didn't even know about),
this is black magic to most users. They need to be told, and if not
understanding, at least informed by names they know what such a shift
will mean. 

It's not going to matter to some. But if it really didn't matter to you,
why not set up a PC with Linux today? 

Dave Haynie  | V.P. Technology, Met@box AG |  http://www.metabox.de