[9fans] hardware documentation (was Plan 9 future)

[9fans] hardware documentation (was Plan 9 future)

[jmk]

	>example, a company who makes another OS (creativly named the
	>BeOS) originally wrote their product for PowerPC (first for

Actually, the original Be machine was a dual-processor Hobbit.

[9fans] hardware documentation (was Plan 9 future)

[Douglas]

presotto@plan9.bell-labs.com wrote:
> Win LT modems are finding their way into most notebooks and
> laptops that I'm interested in buying.  ...
> Hopefully, I'll find the time before the release,
> especially since my favorite IBM laptop has a Win LT modem.

Good luck, but also I think we should be complaining to the
notebook makers that we don't appreciate them on-loading
(opposite of off-loading) such processing onto the CPU.
Throughout the history of computing, we learned to go in
the opposite direction (communication processor frint ends).
This is especially important now that modems are using
sophisticated ECC and modulation techniques -- sure, a lot
of it can be done by the CPU, if necessary, but that's not
what the CPU is best used for.  It's also simply not good
design to introduce so much unnecessary functional coupling.

[9fans] hardware documentation (was Plan 9 future)

[Anthony]

//The move towards a common bus (PCI) may help, but only if we
//have at least one viable competitor to Microsoft

even horridly ugly things can be useful. Microsoft's crap's
always been so bad that anything that could run the latest
version of Windows was entry-level server hardware if given
a real OS. and Linux is useful for getting documentation out
of people, even if it's only in the form of source code.

i've worked with boards from moto, intel, and mips, and have
been impressed with their documentation (yes, even intel's).
but most of these guys make boards, not systems. and turning
one into the other isn't where i want to spend my time, and
certainly isn't my area of expertise.
: anothy;

[9fans] hardware documentation (was Plan 9 future)

[Digby]

> //...the same factors should be making makers of other boxes
> //based on non x86 architectures that much more helpful.
>
> it "should", yes. unfortunatly, this doesn't seem to be the
> case. Apple and Sun, the only two i know anything about, are
> both as dificult as ever. Apple more so, actually, now that
> they've basically killed off the clone market. Be, for
> example, a company who makes another OS (creativly named the
> BeOS) originally wrote their product for PowerPC (first for
> custom hardware, then a port to most popular Macs). they've
> ported to the x86, and are unable to produce a version of
> their OS for the G3 or G4 Macs because Apple won't tell Be
> anything about the hardware.
>
> i hear/read that similar things are true of SGI. luckily, the
> introduction of PCI into Sun boxes is making this a little
> less of a problem. also, with Sun releasing source to their
> OS, this problem could be made much less of an issue for Sun
> boxes. i'd love to be able to put Plan 9 on my Ultra 5...

The common denominator here is that hardware manufacturers
that also sell their own operating system have no incentive
to divulge hardware details. Especially if they are counting
on software sales for the bulk of their profits.

The best bet is to buy boards from company's like Motorola,
who rely on third party OS suppliers and hence produce very
good hardware documentation.

The move towards a common bus (PCI) may help, but only if we
have at least one viable competitor to Microsoft, so that
the hardware makers will not be able to automatically assume
that writing one driver will cover 99% of the market.
Otherwise it may actually be a problem, because whereas
boards currently made for the VME bus come with good documentation
because there is no dominant operating system, is VME is
replaced by PCI, the board makers are suddenly presented
a hugely dominant OS and they may decide the commercial
imperative is to provide a driver for that rather than do
the documentation. Of all the operating system vendors, Microsoft
is the one that will not only fail to write a driver for you,
but will make you pay for the privelege of supporting their
system.

Regards,
DigbyT
--
Digby R. S. Tarvin                                              digbyt@acm.org
http://www.cthulhu.dircon.co.uk

[9fans] hardware documentation (was Plan 9 future)

[Digby]

> //...the same factors should be making makers of other boxes
> //based on non x86 architectures that much more helpful.
>
> it "should", yes. unfortunatly, this doesn't seem to be the
> case. Apple and Sun, the only two i know anything about, are
> both as dificult as ever. Apple more so, actually, now that
> they've basically killed off the clone market. Be, for
> example, a company who makes another OS (creativly named the
> BeOS) originally wrote their product for PowerPC (first for
> custom hardware, then a port to most popular Macs). they've
> ported to the x86, and are unable to produce a version of
> their OS for the G3 or G4 Macs because Apple won't tell Be
> anything about the hardware.
>
> i hear/read that similar things are true of SGI. luckily, the
> introduction of PCI into Sun boxes is making this a little
> less of a problem. also, with Sun releasing source to their
> OS, this problem could be made much less of an issue for Sun
> boxes. i'd love to be able to put Plan 9 on my Ultra 5...

The common denominator here is that hardware manufacturers
that also sell their own operating system have no incentive
to divulge hardware details. Especially if they are counting
on software sales for the bulk of their profits.

The best bet is to buy boards from company's like Motorola,
who rely on third party OS suppliers and hence produce very
good hardware documentation.

The move towards a common bus (PCI) may help, but only if we
have at least one viable competitor to Microsoft, so that
the hardware makers will not be able to automatically assume
that writing one driver will cover 99% of the market.
Otherwise it may actually be a problem, because whereas
boards currently made for the VME bus come with good documentation
because there is no dominant operating system, is VME is
replaced by PCI, the board makers are suddenly presented
a hugely dominant OS and they may decide the commercial
imperative is to provide a driver for that rather than do
the documentation. Of all the operating system vendors, Microsoft
is the one that will not only fail to write a driver for you,
but will make you pay for the privelege of supporting their
system.

Regards,
DigbyT
--
Digby R. S. Tarvin                                              digbyt@acm.org
http://www.cthulhu.dircon.co.uk

[9fans] hardware documentation (was Plan 9 future)

[geoff]

Mr. Choat wrote:
> Exactly, the reason it was 'open' then was there was no real competition.

No, hardware documentation was `open' because there was competition
among peripherals vendors and good documentation was a competitive
advantage (there's no need to reveal all details of a controller's
implementation, merely the interface to it and enough
theory-of-operation to drive it).

> The licensing agreements were still there. If you don't believe it then
> try to explain the comments on pp. xi or the back cover of,
> Lion's Commentary on Unix, 6th ed.

These comments all refer to software (Unix) licences; they have
nothing to do with (non-existent) hardware documentation licences nor
fees for same.  I wrote Unix device drivers before working for Bell
Labs.  I don't recall ever encountering (or even hearing from others
about) licence fees for hardware documentation.  Perhaps when hardware
documentation got scarce later, somebody took a bribe for providing it
illicitly and called the bribe a `licence fee'?  :-)

[9fans] hardware documentation (was Plan 9 future)

[Anthony]

ah, right. i had forgotten about that
generation. they had moved on to the
PowerPC by the time i got involved. i've
still got my dual-603e sitting about 3
feet away. now if only there was a Plan
9 or Inferno port i could get for it...
: anothy;

[9fans] hardware documentation (was Plan 9 future)

[Anthony]

//...the same factors should be making makers of other boxes
//based on non x86 architectures that much more helpful.

it "should", yes. unfortunatly, this doesn't seem to be the
case. Apple and Sun, the only two i know anything about, are
both as dificult as ever. Apple more so, actually, now that
they've basically killed off the clone market. Be, for
example, a company who makes another OS (creativly named the
BeOS) originally wrote their product for PowerPC (first for
custom hardware, then a port to most popular Macs). they've
ported to the x86, and are unable to produce a version of
their OS for the G3 or G4 Macs because Apple won't tell Be
anything about the hardware.

i hear/read that similar things are true of SGI. luckily, the
introduction of PCI into Sun boxes is making this a little
less of a problem. also, with Sun releasing source to their
OS, this problem could be made much less of an issue for Sun
boxes. i'd love to be able to put Plan 9 on my Ultra 5...
: anothy;

[9fans] hardware documentation (was Plan 9 future)

[Jim]

On Fri, 12 May 2000 geoff@x.bell-labs.com wrote:

> In fact, before Microsoft's dominance and ability to bully
> manufacturers, it was possible, indeed easy, to acquire programmers'
> documentation for peripherals, because it was in the manufacturers'
> best interest to make it freely available.  And nowadays ``Linux'' is
> the ``Open Sesame'' that sometimes works to dislodge hardware
> documentation, though that's due to politics not technological
> excellence of Linux.

Exactly, the reason it was 'open' then was there was no real competition.

The licensing agreements were still there. If you don't believe it then
try to explain the comments on pp. xi or the back cover of,

Lion's Commentary on Unix, 6th ed.
John Lions
ISBN 1-57398-013-7

    ____________________________________________________________________

            The future is downloading. Can you hear the impact?

                                        O[rphan] D[rift>]
                                        Cyber Positive

       The Armadillo Group       ,::////;::-.          James Choate
       Austin, Tx               /:'///// ``::>/|/      ravage@ssz.com
       www.ssz.com            .',  ||||    `/( e\      512-451-7087
                           -====~~mm-'`-```-mm --'-
    --------------------------------------------------------------------

[9fans] hardware documentation (was Plan 9 future)

[presotto]

It is true that Microsoft's dominance has has an influence on
the nonexistence of hardware documentation, though I think its
not so much because of bullying.  By making the x86 architecture
the only one worth building hardware for, Microsoft turned
computing into a commodoties market.  That brought the price of
hardware down.  More signiicantly, the manufacturer's margin and
time to market came down and differentiation between products
became negligible.  Hence, there is nothing driving a company
to disclose hardware info:  one driver gets them 97% of the market,
documenting mistakes is embarassing, and disclosing too much
information helps your competitors copy or surpass you.

Of course, the same factors should be making makers of other
boxes based on non x86 architectures that much more helpful.
I've been told Apple is a lot easier to get info out of these
days though I personally have not experienced such.  The reduced
time to market makes everyone harder to get info out of because
they really don't have it encapsulatd in a usable form.  In
many cases we've gotten Unix, Plan 9, and Linux drivers for
new hardware only because someone inside the company was enough
of a fan to take the time to help get the driver built in
his/her spare time.  In most cases the company wasn't being
paranoid or secretive, it just didn't have the time to devote.

Of course, I'm now seeing secretive from the inside out.
Win LT modems are finding their way into most notebooks and
laptops that I'm interested in buying.  Lucent spits the
work between the chip and the host cpu and is very afraid
of letting the code out.  There was plenty of time and money
to get a Windows driver out since it represented a huge
market.  Due to Linux fans both in Research and Microelectronics,
a binary only Linux driver now exists, with he company
actually supporting it.  I've just gotten the source code
for it with the stipulation that I also release binary only
versions for Plan 9 (if and when I figure out the code).
Hopefully, I'll find the time before the release,
especially since my favorite IBM laptop has a Win LT modem.

[9fans] hardware documentation (was Plan 9 future)

[geoff]

Mr. Choate writes:
> Again misreprensentation.  [...] Second, BSD has never had as large a
> community and if BDS got access to hardware it was because somebody
> paid the vendors licenseing fee (usualy Berkley or a derived
> authority), it was not like with Linux where the vendor releases the
> info Open Source.

In fact, before Microsoft's dominance and ability to bully
manufacturers, it was possible, indeed easy, to acquire programmers'
documentation for peripherals, because it was in the manufacturers'
best interest to make it freely available.  And nowadays ``Linux'' is
the ``Open Sesame'' that sometimes works to dislodge hardware
documentation, though that's due to politics not technological
excellence of Linux.