Re: Oh why did Apple dump IBM....

Elliott Roper <nospam@xxxxxxxxx> wrote:

Rowland McDonnell <real-address-in-sig@xxxxxxxxxxxxxxx> wrote:

Elliott Roper <nospam@xxxxxxxxx> wrote:

Rowland McDonnell <real-address-in-sig@xxxxxxxxxxxxxxx> wrote:

Elliott Roper <nospam@xxxxxxxxx> wrote:
<snip>
X86 will die eventually when compiler back-ends finally do what they
are supposed to do.

What *are* they supposed to do, then?

I can't see why you could not make a machine with a
completely arbitrary ISA, made sufficiently mutable for the task in
hand.

Writing compilers would be a pain, surely?
The back end is supposed to insulate the hardware peculiarities from
the language peculiarities. If the backend were doing its job well, it
would be easy to lurch from one ISA to another, or even create an ISA
that suited the problem. A kinda back back end.

I think you think that ISA means something other than what I mean by
ISA.
I'm using it for the instruction set visible to an assembler programmer

And I'm talking about the CPU's instruction set visible to the assembler
programmer - you seem to want to include the ISAs of all the other bits
of the computer.

I don't see how it's possible for a compiler to do what I think you say
it ought to do. Hardware peculiarities are just that - how could a
programming language be expected to `know' in any form at all about the
particular features of the particular (say) fancy graphics processing
hardware built into the original Amiga? Seems to me that the programmer
has to learn about these features and work them himself directly.
You missed the bit where I made a distinction between the front end and
the back end of a compiler. The back-end guys still have a lot of work
to do. With the right programmable microcontroller and FPGA *beneath*
one ISA, it might not be impossible to present a mutable ISA.

Umm... Hang on a bit - the ISA is what the compiler writes, yes? So if
you're talking about a mutable ISA - well, it seems to me that you're
talking about designing a computer that can modify its ISA on the fly to
match whatever the compiler kicks out so that the hardware can adapt to
whatever the compiler demands of it. Which doesn't seem like the right
way to go about things to me.

It was
done in a small way with so called PAL code on Alpha. It gave Alpha
some VAX-like instructions that the VMS internals people needed badly
for spinlocks and queue/dequeue that had to be atomic.

I'm really not at all sure I know what you're talking about now. You do
seem to be talking about extending the ISA of a particular CPU - well,
why not? And if you do that, of course you modify the compiler so that
the new instructions can be used to compile your high level language.

But I don't see what that's got to do with programming in assembly or
/future advances/ in compiler back ends - have I missed something?

Once you've got a pukka ISA defined for your graphics hardware, it's a
different matter - but I've been referring to the CPU's ISA, not
anything else.
..not an important distinction in principle. Watch some clever math
being done in Playstation graphics processors.

An important distinction in principle because the `programs' (OS, apps,
all that jazz) that you have installed on your computer can't run using
the graphics processor in your graphics card to stand in for the CPU it
was designed to run with.

You're locked in to using your CPU for that. For sure it is not
something that need necessarily be the case in principle - but with what
we've got, that's the way it is.

That sad little x86 mess has survived so well because the hardware
has got so quick so quickly,

No, it survived because IBM decided to use it for its PC, and then
MS acted illegally to wipe out the competition, leaving us with an
ecosystem dominated by - well, I'd call it `***', but if you pile
*** up and leave it rot, it's damned useful stuff.

Nope. It is a nice conspiracy theory, but the facts don't support it.
The first instance of Windows NT was written to run on x86, Alpha and
some MIPS thing. Later it ran on Itanic as well. It still does.

<puzzled> I fail to see the relevance of your point - all that happened
long after MS had created the x86/MS-DOS world for us using the
well-documented tactics that I described. And there's no doubt from
reading the facts that what happened was in fact what I say did happen.
Well, I didn't appreciate how finely your argument hung on a particular
early period of the x86's life.

<puzzled> Well, I did specify that that was what I was talking about -
it only happened once, and that was when it happened. What I was
talking about was just that part of its life - everything that's
happened to x86 since the `great extinction' of the 1980s has been a
consequence of that cataclysmic event which you don't seem to want to
pay any attention to at all.

One cannot talk about the world of business computing these days and
ignore the earlier events - which were aeons ago in terms of computing
developments, but not all that long ago in other terms. But you seem to
just shy away from the events that set the scene we now see - why won't
you pay attention to them?

It's like trying to talk about modern German history without mentioning
Hitler and the Nazis (and no, I've not lost: I've not compared anyone to
either of those, have I?) - it's there in the background, this big
lurking horror.

You did not mention the several
Motorola based processors that were developed in the same timescale.

That's right - none of them made it. All were swept away. For sure the
PPC line is doing very well in some areas, such as embedded controllers,
but they're gone from the PC market pretty much completely, aren't they?

They ranged from the incredibly cheap 6800, the slightly nicer 6809,
and the 680x0 family, which competed fairly well with x86 for a while.

But it fell by the wayside - come the late 1980s, it was obvious that
the IBM Compatible/MS-DOS combination was taking over the world.

By then, we had IBM compatible PCs for the suits; Amigas for video bods,
the whole MIDI thang, and hip games players; Ataris for Idunnowhat; and
Macs for `creative types' and who got sneered at by business people who
thought they knew about computers (I spoke to some - quite astonishing,
the violent arguments they came out with to explain what was crap about
Macs and how it was MS-DOS was perfect for business needs. And you
could provoke the torrent just by mentioning that you quite liked the
idea of a Mac; I didn't meet one in real life until 1990). Arcs were -
and I felt sad when I saw the way it had to go - on their way out due to
having no niche at all.

Ataris went sooner, Amigas went later. Arcs hung on longer than I
thought possible. The Amiga and Archimedes platforms aren't quite
entirely expired, mind. But it's hard to tell at times...

The 680x0 made Sun Microsystems and, at least as far as I remember it,
spawned the whole low cost unix industry. I could have mentioned it was
a major factor in the fall of DEC.

The 68000 line was certainly competition for DEC in that sense, yes.
But it's the PC market that x86 and MS took over, wasn't it? DEC wasn't
competing there, was it?

68000s were always more significant in the heavier duty machinery. I
recall using a 68020 powered Apollo DN330 - with its 70MB HDD and all
and a massive 1MB RAM. It still counts as a massive 1MB RAM - modern
computers have much less RAM on the whole. Yeah, that's right: you
should have seen the size of the board, covered all over on both sides
with chips as I recall, about 18" on a side.

And I really do object to you calling my idea a conspiracy theory -
that's just gratuitous libel to discredit me. I posit no conspiracy.
The only reason for accusing somone of a conspiracy theory is to
discredit their ideas using dishonest tactics and I'm really quite
pissed off with you for doing so. btw, conspiracies are commonplace and
`how the world works'. Unlike conspiracy theorists, I make very few
claims about what's going on with any of the myriad conspiracies around
us. So please do drop that line, eh?
I wouldn't have given you credit for an argument that was well
established as a conspiracy theory before you claimed it for your own,
so you may conclude there was no personal insult intended.

<puzzled> It's not so much a conspiracy theory so much as `what really
happened' as far as I can tell.

Conspiracy theory to my mind means `Pile of paranoid ravings based on
the flimsiest of evidence and constructed by someone with no idea about
rational thought' so I think you're being quite insulting.

The point about conspiracies is that they're, well, conspiracies, so any
competent conspiracy will be hard to find out about - so anyone
investigating serious conspiracies will get a bit potty. I have heard,
mind, that one of these researchers into odd stuff is used by CIA staff
to help them find out what the bloody hell is going on inside the CIA.
I'm not sure that she's necessarily right, but I get the idea that she
finds out enough pointers that those on the inside can detangle things
adequately.

But you know, Conrad Black's just received a 6.5 year prison sentence
(oh happy day!) for activities he conspired to enact - conspiracies are
all around us. They're real. Hitler, for example, conspired to take
over half the world and kill all the Jews. Most people are a bit less
mad with their conspiracies - some people just conspire to get some
biscuits from the kitchen when mum's not looking.

The fact of the matter is that x86 performed well at a good price
point.

The fact of the matter is that the benchmarks I read for the first IBM
PCs showed that they were slower than BBC Micros - 2MHz 8 bit beat
4.77MHz 16 bit (running interpreted Basic, please note, so it's a bit of
an iffy comparison). So I think the claim that they performed well is a
bit much given the evidence. IBM PCs had no graphics abilities either
(unless you really want to count the CGA thing as graphics). They were
in terms of computing power per pound spent /very/ poor value for money
compared to the competition back in the early days - slower than the
competition, much less well equipped except for the lovely keyboard and
solid case, and were a lot more expensive. So you're wrong when you say
they performed well at a good price point - the reality was the opposite
of that.

What they did have was the IBM label, a lovely keyboard, a good solid
case that business people felt they could trust, and Lotus 1-2-3.

As competitors like Motorola improved their more elegant
machines, Intel time and time again made the wretched Pentiums go
faster and faster.

Pentiums? Why are you talking about Pentiums? Pentiums came out years
after the events I'm referring to. MS had done the job of wrecking the
old PC ecosytem years before Pentiums appeared.

Much of their price advantage was the awesome scale
of their production.

Yes indeed - but the original Pentium was a bit crap really. Without
code optimised for it, you were never going to make good use of the dual
processing core arrangement. They did blow 68040s completely into the
weeds, mind (well, 486s were doing that anyway...); but put up a poor
showing against the PPCs when they came out.

They were able to make complicated cruddy machines
with insane fetch and execution pipelines far cheaper than Motorola or
Silicon Graphics or MIPS or DEC could do. Intel saw the lot of them
off. It had little to do with Microsoft, and nothing to do with IBM,
who almost from the start, were not that fond of Intel in any form.

But that has nothing whatever to do with the processes I was talking
about.

The fact is that Intel and AMD managed to keep the real world price
performance of that pile of *** ahead of more elegant processors.

Of course - by the time WinNT came out, x86 had destroyed the
competition: there was no way any competitor could have got the sales to
match x86 development funding by that point. It took a triple alliance
between Motorola, IBM, and Apple to create the PPC line from two
existing architectures to put up any sort of competition to x86, and
while IBM's POWER architecture is still going strong in servers, PPC is
now dead on the desktop.
Nope. You have the timing a little bit wrong

I think you're missing the point I'm trying to make.

You might remember that
Intel was running scared of Alpha at that time.

I never met a Windoze box running on Alpha, so I don't think Intel can
have been very worried about Alpha on the desktop.

Itanic was going to be
saviour of Intel.

Well, yes.

Itanic flopped because good as it was, the race
between Intel and AMD in the x86 arena continued, against all
expectations, to produce faster cheaper cooler running machines.

That's one interpretation of events. Another reason might be that it
wasn't /that/ good, had some fundamental problems due to dodgy design
decisions, and Intel didn't put the effort into developing it that it's
put into the x86 line.

Intel's famous for taking the wrong route - look at its insistence on
`clock speed at all costs' for its PC CPUs leading it astray until very
recently. Everyone else was going for wider and multi-thingy, but Intel
was just trying to ramp up the clock rate. Oops! But how long did it
take Intel to recover from this mistake once it decided it was going to
go in the right direction? Not long - it was just a matter of `doing
what everyone else is doing' and shovelling on the development
resources. Intel caught up quickly because it could put huge resources
onto the job.

In the case of Itanium (or whatever we're supposed to call it), Intel
had to make the right decision in-house rather than `do what everyone
else is doing' - and just happened to get it wrong, or so it seems to
me. Sometimes Intel gets it right, but frankly most of its designs that
I've looked at seem a touch inelegant to me. Okay, I've not actually
looked at many Intel chips in any detail, but I've always been slightly
nauseous when I've done so.

<shrug> It's possible I've got a bias against Intel, but I really do
think it's something that comes from my personal sense of aesthetics not
matching the Intel corporate culture which causes its designs to come
out the way they do.

History shows us that MS did in fact create an ecosystem where nothing
else could survive other than the x86 line of CPUs.

Microsoft would have dumped x86 if Alpha had ever delivered on its
early promise. It didn't and DEC died. They bet the farm on it. There
was some dirty work at the crossroads, from Intel rather than
Microsoft, but by then DEC had had it.

How is this contradicting what I said? I see no contradiction.
Microsoft used what was best for their business.

Well, yes - what was best for their business was using illegal tactics
to wipe out the competition. They did that, Intel gained big time, and
then all this business with Alpha and whatnot turned up.

They would have dumped
x86 if it failed to improve. Therefore it is incorrect to assert that
Microsoft acted illegally to wipe out non x86 processors.

What I have asserted is that MS acted illegally to wipe out the
/software/ competition, and that had the effect of wiping out the
processor and PC platform competition too.

Microsoft has used sharp business practices more than once, but the
courts in the US and Europe never pursued Microsoft for establishing
its monopoly in the first place. That was not illegal.

I've read that they've been done for using illegal practises to set up
its monopoly - but I'm damned if I can be bothered to check the details
now.

They pursued
Microsoft for leveraging its monopoly into new markets (in the case of
the EU: workgroup file and print servers)

You wouldn't care to translate that into English, would you? I've
worked out that `leveraging' in that sort of sense usually means
`using', but that doesn't make sense in this case.

Why can't people just speak English instead of using these crappy
bull*** buzz-words that don't actually mean a damned thing?

DEC sued Intel for some IP
infringement. Intel "lost", got all DEC's compiler team, the Hudson
chip fab, paid DEC $800 million and agreed to fab Alphas. It bought DEC
enough time to be eaten by Compaq, who found it indigestible and was
itself eaten by a printer ink company. I don't remember its name. I
believe it once made quite reasonable lab instruments but long ago sold
that business off.

Righto.

there has been little economic advantage in
doing otherwise.

<puzzled> But there would have been huge economic advantage to any firm
that needs computers to be used by people if they'd not invested in MS
operating systems. The unreliability of Windoze in particular, and the
costs of dealing with it must rack up annual global costs far greater
than the GDP of most nations.

Possibly true. I'm inclined to agree with you, mostly from shared
prejudice,

It's not prejudice in my case. It comes from looking at the internal
architectures of the various CPUs and deciding what I liked and so on.
I always like 6502s. I always found Z80s less elegant.
As your discussion with Richard Tobin has illustrated, beauty is in the
eye of the beholder.

He prefers Z80s because he likes the ease of programming with the
instruction set. I prefer 6502s because they're stripped down dirt
cheap speed monsters.

It all depends on your point of view - but to take your rather sneering
tone and ignore the details of why we each have our preferences is
rather insulting to my mind.

The first PC I had (well, 'twas actually my dad's officially - 1981, you
know?) was a ZX81 and I'll say now it's a lovely piece of work for what
it is. But I prefer the 6502 approach to the Z80 one.

The Z80 was an 8080 with one extra index register and a different
assembler syntax.

<cough> /And/ an entire duplicate register set - which could be very
handy if you thought to use it. Does the 8080 do DRAM refresh as part
of the machine cycle? Has it got the same nice i/o and control line
architecture designed to work with that coherently designed peripheral
chipset the Z80 had, and that made the Z80 such a piece of piss to
design into things?

FWIW, I hated the 6502 nearly as much as the Intel
chips of the day.

Why?

I preferred the 6809 and later the 68000. They looked like sawn-off
11's to me, and I was happy with that.

Righto.

Anyway, it struck me long before I'd heard of IBM PCs or RISC or
anything that RISC beat CISC. And then I studied - as part of a lecture
course - the 80286 vs. the 68020. The 80286 internal architecture just
made me want to vomit, it's just so *nasty* by comparison with the
elegance of the 68020's internal arrangements. I don't recall if it
still had the 8086s segment registers. I suppose it must have, but I
think I must have been deliberately trying to forget the full horror of
the details.
I'll heartily agree with that. I'm pleased you mentioned the 68020
unprompted. It shows that the x86 had not completely won in your
preferred timescale of discussion.

But it had - by the time the 68020 was current in PCs, it was obvious to
everyone with any wit that the PC war was going to be won by the
MS-DOS/IBM side, although the fighting wasn't yet over. Almost all PCs
in office use were x86 powered by then - that's what did it. Arcs had
just been born, but were clearly going to die soon; I for one was
astonished at how long they lasted.

No prejudice: the x86 series of CPUs is a *MESS*.

but the ISA of the machine underneath did not matter.

I'm not so sure about that. The Wintel world is inelegant from the CPU
upwards. The fact that IBM chose an inelegant CPU at the start, used an
inelegant internal hardware architecture, and

Economics plays much less part in decision making than economists have
been wont to claim - people ignore the costs associated with Windoze, so
I've noticed.

I once worked for a firm that decided to get a network of Windoze
computers in. It never did work properly. First day, printing packed
up - it had allegedly been working on installation over the weekend. I
got asked to fix it. I pointed out that 1) I was employed to write and
2) I told 'em to get Macs 'cos Windoze would be nothing but trouble so
don't expect me to do anything but smirk at this point.

I might have done the same. History shows we were wrong.

But it has done no such thing.
Come on Rowland. A bit of consistency if you *don't* mind.

<puzzled> But I am being perfectly consistent.

Just up the
page you asserted that x86 *did* wipe out the competition.

What I've tried to say is that the x86 came to dominate the PC
marketplace due to MS-DOS dominating the PC marketplace and also the IBM
PC-compatible type of MS-DOS PC being the only one that people wanted
for software reasons.

And what's any of that got to do with Windoze being crap? Also, there's
no contradiction between saying that the early x86s CPUs were crap *AND*
that the software that ran on them wiped out the competition.

After all, even a crap CPU can be bloody useful. I'd just rather the
nicer CPUs had ended up on top. Not that you'll hear me complaining
about the performance of the x86 line in recent years, mind. But note:
performance, and recent years.

I backed the
wrong horse. So, it would appear, did you.

But I didn't - they had terrible trouble with that network and the
software. Lost data, cursing, swearing, printer not working, and so on.
The staff concerned had in fact been happier and much more productive
with the Amstrad PC clones that they'd just had replaced which had been
running Wordstar 2000 under some version of MS-DOS.

I'd told 'em, I had: I told 'em what the trouble they were going to get
was, and the trouble I predicted turned up. `It was just trouble',
basically. At least I managed to save them from WordPerfect...

I'll say lots of harsh things about MS-DOS, but it had the great virtue
that if you set up an MS-DOS box to do a job, it would do the job to
spec, reliably. But then, that's because it wasn't designed by MS.

The whole business world has
decided that Windows is better value for money than much better
engineered competition.

Yes, that's right. But what is it that makes them decide that?

How did that come about? You need to look at the world before MS
Windoze arrived. And do bear in mind that hardly anyone in business has
ever done a cost comparison analyse of total lifetime costs of a
Windoze-based network and anything else so while they've made a
decision, it's not actually a rational decison based on facts: it's just
prejudice. Or it almost always was back in the 1980s - I got to find
out how it worked in real life on a number of occasions. It was all
based on prejudices of the dominant players in the `what sort of
computer are we going to get' process - essentially, anyone who didn't
suggest IBM PC wasn't going to get the supply contract back then unless
there was some sort of link to publishing. Facts were never
/considered/, just presented for consideration so that those which
backed up the decision could be selected to justify the decision after
it had been made. I gather that this is how most business decisions are
made: it's not very rational, and I wouldn't trust a normal business
manager to run anything.

What they mostly do is `find out what everyone else is going and then
copy them' - which is what happened, in part.

Back in the 1980s, starting in 1981, business world decided that it made
more sense to drop a few grand on an hugely expensive and rather
underpowered IBM PC running a 16 bit rip-off of an 8-bit 1970s OS rather
than the much cheaper and mostly just as capable competition.

Why, I cannot say - but the fact that `no-one go sacked buying IBM' is
one reason, as is the fact that it looked more businesslike than the
competition.

Anyway, they made stupid decisions back then, but they weren't so stupid
as to damage the firms (these machines were after all reliable, if
expensive and slow - but those downsides didn't matter at all in office
use), and they got it to work to what turned out to be a useful extent.
After all, even the distinctly 1970s tech of the original IBM PC could
improve an old-fashioned office quite a lot, couldn't it? Lotus 1-2-3,
word processing, database, mail merge: all these things were very, very
handy for business people.

Now, after that lot of decision making had pushed the business world
towards MS rather a lot, MS came out with MS Win and suddenly these
businesses could get a fancy GUI that apparently would increase
productivity and so on without much extra investment at all: no need to
replace your computer hardware, no need to replace any of your software.
Just add Win 3 and a mouse for instant GUI happiness!

Anyone with an MS-DOS box at home acquired Win 3 via the installer from
work and so Win 3 took over the world on the back of MS-DOS's
considerable market pentration and deliberate refusal to assert its
copyright for early versions of various bits of software in order to
achieve high installation rates through software piracy (MS has admitted
this is one of their tactics, by the way - I've read it in a quotation
from a senior MS bod; that's an admitted conspiracy, so not any kind of
conspiracy theory). Various software lock-in tactics, natural
conservatism (i.e., a desire to resist change), and so on did the rest.
Once MS had `everyone' (the office world, that is) using Windoze, it
then started to get a bit tighter about charging for it, you might have
noticed.

That's how it happened - nothing to do with deciding what works best,
just a series of ad-hoc decisons taken at the time that resulted in
certain outcomes. It was managed in part by MS which did do long-term
thinking and was run by Bill Gates, who's a very canny chap and knows
how to manipulate the business world. No, I don't say this was all some
sort of take-over-the-world Machiavellian cunning plan, just normal
businessmen doing the normal job of normally playing their normal market
in the normal way to use normal methods to get the biggest market share
as they can - it's what you'd expect any hugely successful businessman
with lawyers in his background - Gates is one such - to do (i.e., break
the law in devious fashions to make more money from business
activities).

I don't see what makes you so sure that business decisions are always
right or even optimised at all - why shouldn't business people make
decisions that are initially stupid but then get them to work? That is
in fact the way business works. Sometimes the decisions are worse
optimised than others; they're rarely the best possible decisions.
After all, they're all made by people, and mostly these people are
incapable of making rational decisions even if they have the relevant
facts, which they usually don't. And don't tell me that's not how it
works: I've met enough of it myself.

Now, once you've got a critical mass of `people working one way with one
set of tools designed to force lock-in', you can't easily change - given
the difficulty of getting it all to work in the first place and the cost
of changing, it became enormously difficult to change to anything else.

And of course if `everyone' is doing it your way, it's no big commercial
disadvantage if you do what they're doing. So for all that the Windoze
way is sub-optimum, it functions adequately for businesses, and you
don't get a big commercial advantage by doing things another way. So it
goes. That does not mean that Windoze is best in any way at all - just
that you can make good money using it.

The fact is that well before MS released Win 3, we already had lock-in
and the performance of the actual computers and software in itself
stopped mattering: what came to matter was what you could expect your
staff to use with no training, and what you could expect to work with
your clients, suppliers, partners, existing software installers,
existing data, and archives. Etc.

Do bear in mind that good CPU performance is not something that a
business needs for its office computers on the whole.

I had a boss at DEC who would ask "If you are so smart- why aren't you
rich? Another at another company who said "Money isn't everything, but
how else do you keep score?

I know that's the sort of thing that these people say. What of it? It
doesn't stop them making stupid decisions - and as I've pointed out,
well before Win 3 came out, the PC ecosystem was such that it *would*
have been daft for most business people to have bought `other than
Windoze' from many points of view.

But I was working for a publishing firm - the big downside of Macs (poor
interoperability with MS-run PCs of the era) isn't an issue inside the
publishing world. Quite the opposite, if anything.

Economics did
and does play a leading role. For many organisations, Windows is great
value for money.

But that has not been measured.
Well, everyone with chequebooks measured it to their satisfaction I
guess.

MS Win 3 famously got to `critical mass' by being installed on
`everyone's home PC using discs from the office'. So yes, people
measured with their cheque books: I can get this for free, is how it
worked. The businesses bought Win 3 because it was the cheapest way to
get a GUI: they didn't have to replace any software, they didn't have to
replace any hardware: all they had to do was add one bit of software.
Cheap, you see - due to the existing MS lock-in that had been
established *before Win 3 came out*.

It all goes back to that - once Win 3 had taken over, that was *IT*.
The writing was on the wall in the 1980s with the great extinction - but
Win 3 put the lid on it.

Better value than MVS, VMS or OS X. I don't understand
it, but they are successful, and I'm not.
That is, if nothing else, economics working as designed.

Erm?

You can succeed when you have a poor solution in place - what makes you
think that the converse is true? And people do succeed with these poor
solutions.
You and I might think they are poor solutions. They did succeed in
money terms, therefore they are not poor solutions, in spite of our
better taste and judgement.

Argh. No, sorry, that doesn't work. It's perfectly possible to make a
good living selling a substandard product. Many people are willing to
buy substandard products due to low price. Many people insist on
substandard products for other reasons.

MS-DOS was unquestionably a sub-standard product, being little more than
a 16 bit rip-off of the 1970s PC OS, CP/M. But it won!

How come?

Well, a poor product can still do the job - if it does the job, it's
good enough, right? Don't forget that the business people who make the
decisions on what computer platform to standardize on generally have no
way of judging what's best other than `what does everyone else do?'

If it does what's expected of it by the buyer, then it's `good enough'
regardless of whether or not it's better or worse than the competition.
And if they buyer knows *** about the competition - as was almost
always the case - he's not even making a rational buying decision, but
such was the usual process.

So: all you need is for `whatever it is' to be *good enough*. MS
Windoze is good enough for business use (not really, but they've got it
to work in a fashion that functions for some values of `functions'; I
have my doubts when I meet people in shops swearing at their Windoze
boxes and holding things up in the shop 'cos of crappy MS crappy
software, mind) - that's all that's required. The rest of it's
marketing, and MS has been very clever with that by making sure
manufacturers only sold computers with MS-DOS and no other OS, using
dodgy tactics like deliberately encouraging software piracy to achieve
dominance in the installed base (leading to dominance in the marketplace
as the next step, at which point you get tighter on making sure people
pay), and so on.

Oh, it's enormously complicated, all the things that went on - there are
many other factors, such as (for example), Bill Gates knowing exactly
how to sell to business people, while The Steve has tended to wind them
up. It goes on like this.

Okay, early IBM PCs did have a big attraction back in the early days:
Lotus 1-2-3. But aside from that, they've never had any inherent
technical advantage over the competition, and it was only after they'd
achieved market dominance that the prices started to drop down to
slightly more reasonable levels. I must repeat this: in the early days,
IBM PCs were *HORRIBLY* expensive compared to the competion: they were
not, as you claim, remotely cheap. Only Apple thought to sell at a
higher price come 1984 so of course hardly anyone bought a Mac - after
all, who could afford one?

This came about in part because US computer firms priced computers for
the UK market by putting a pounds sign where they used to have a dollar
sign, so we paid nearly twice as much for US computers as we should have
done. Apple ][s were always rather pricey for this reason.

There's nothing `great value for money' about a secretarial computer
network that needs repairs every week, that doesn't print when you want
it to, and that loses data. All this is commonplace with Windoze and is
what I saw with the system that had been installed in the above firm.
Well, Apple and lots of little companies could not compete. Windows
customers were not taking bribes or anything.

But the firms that made and sold MS-DOS computers were succumbing to
illegal commercial pressure: they could get MS-DOS at a discount price
if they only sold computers running MS-DOS. By the time Win 3 came out,
that had been stopped but the damage to the PC ecosystem had been done.

I would say that it's much poorer value for money than what they'd've
had with Macs (System 7 was recently out) at the time - which I knew
would `just work' when networked to the printer, for example.
Look, I agree at the personal level. If *I* had to use a Windows box,
I'd give up using computers.

That's irrelevant.

In fact I put my money where my mouth was.
I ran my company down the drain trying to sell well engineered reliable
VMS systems on VAX and Alpha instead of somehow making it work on
Windows.

That was trying to buck the trend and wasn't likely to work. Just
because I think that all these people buying Windoze got it wrong
doesn't mean that you'd've found me trying to sell Macs - it would
clearly have been a thankless task since by the time Win 3 came out, MS
had won and business people were prejudiced against those who thought
that MS-based computers weren't the optimum solution - essentially, by
trying to sell non-MS based computers, you were cutting your own throat.

Back when I still taught for a living, I spent a year teaching typing.
One class was using Windoze 3.11 for workgroups (IIRC). All networked.
All HD-less netboot machines. Networked printer.

Every single lesson was disrupted by data loss due to filthy floppy disc
drives or viruses - only the latter can be blamed on dodgy architecture.
Almost every lesson was disrupted due to printing problems.

When the boss of that department (I was actually physics at the time)
told me happily that my other class - using Amstrad PCW8512s - was going
to be `upgraded' to Windoze PCs after Christmas, I begged to let me keep
using the Amstrads.

btw, one class all passed. The other class did not. Guess what? Those
using the Amstrads did well. Those who had had to suffer Windoze did
not.

Admittedly, part of the problem was that the tech support was provided
by a priest who was too vile to be let teach (this was an RC 6th form
college). He refused to run virus checking on the HD-based machines on
the network, and refused to clean floppy disc drives.

As a result, the college network was an unreliable plague pit and
disaster zone.
Yes. There are plenty of anecdotes like yours. I have a few too. But
but but. Intel walked off with the money.

Yes I know - and they did it on the back of the crap technology of MS's
operating systems which MS happened to market very well for the benefit
of MS only - but the way it worked out, Intel's manage to catch a ride
off MS's coat tails, regardless of what MS might have liked to have
seen.

You can't complain too much about x86 these days, because they work
pretty well for all that they're inelegant. But Windoze - ARGH!!!

I'm typing this on a G4, but
next door to it is my Mac Pro, cruddy ISA and all, that is at least 10
times faster and the basic box cost less than this poor l'il Powerbook.

Well, yeah? And? So computers get faster - it's a bit silly comparing
an ancient 32 bit single core G4 with a brand spanking new whatever
multicore monster 64 bit CPUs your Mac Pro has fitted to it.

This 'ere 4G5 would probably put up a non-pathetic showing against your
Mac Pro and it totally annihilates the Apple ][s in the cupboard (not
that I've ever powered up any of 'em; don't even have monitors. Just
the main boxes and some disc drives. I just didn't want 'em thrown
away, you know?). But so what?

And you say 10 times faster - well, that might be the case for some CPU
and graphics benchmarks, but it ain't the case for disc access and
probably isn't the case for RAM access to a single CPU, and yer PCI bus
runs at whatever speed PCI bus you've got, sort of thing (or whatever's
designed into the thing, you know what I'm on about).

That is my point. The crud won. The crud won because it did not matter
it was cruddy.

Yes, that's what I said.

It won because it was faster and cheaper. Much faster.

Ah, but you are wrong there. It was much more expensive and it was much
slower but it won. Then it got faster and cheaper than it had been -
still more expensive than all but the most expensive competition (Macs),
but faster on the whole. With the advent of the first PowerMacs, it
lost the speed edge again, but that made no difference at all because by
then the war had been lost.

(Much slower? Yeah, well, compare an IBM PC of the early years to an
Amiga of the same era and then see what I mean - look at all those fancy
CPU support chips in the Amiga. Look at what *they* can do! Whee! and
the Amiga was cheaper, and had a more nearly proper OS)

Much cheaper.

But they were more than ten times the price of plenty of other PCs when
they came out! They were *ENORMOUSLY* more expensive than the
competition, until Macs came out, that is.

It had nothing to do with Microsoft.

But since the IBM line came to dominate in the era when IBM PCs were
very expensive and when MS was famous for using illegal business
practices to increase MS-DOS's market share, I have to say that you're
wrong.

The crud next to me
is running unix with a rather nice GUI. I can't bring myself to even
*think* about programming it in its native assembler.

<shrug> So don't. The only x86 box I have in the house is a broken
Psion 3.

(us editorial types were still using an MS-DOS based system, which did
have the virtue of reliability)

How else did Alpha die? It was RISCy PDP-11 at heart,
yet there was never the money to keep it real-world faster than the
horrid one.

I was under the impression that the problem was selling the thing
because MS had used illegal tactics to destroy the old vibrant PC
ecosystem, thus reducing the scope for different CPUs in the
marketplace.

No. Alpha was never part of a vibrant PC ecosystem.

Erm, what? Alpha turned up after the old vibrant PC ecosystem had been
destroyed. Alpha has nothing to do with what happened *BEFORE* Alpha.
You never made it clear which time you were talking about.

I thought that I had done.

Would that it were.
I bet my company on it and lost. Year after year, there were new Alphas
that were going to swamp the Pentium things. They just never did.

Of course - it was all too late. The damage had been done by then.
Not at all.

But it was, I tell you.

[He's behind you!]

What I'm talking about happened long before then. The war had been lost
by that time. If you think otherwise - well, you're not talking about
what I'm talking about and it's a bit silly of you to carry on saying
that I'm wrong when you're talking about a different thing to me.

x86 ishness improved many many times in performance after
Alpha began production. I repeat, ugly as it is, it was just damn
better at getting the job done.

But by that time, it didn't matter at all.

In
the end Microsoft just gave up compiling Windows for Alpha. DEC was in
deep poo and could not afford either the hardware or software costs to
keep it, or DEC's own operating systems competitive, nor to help fund
Windows development on Alpha.

IN the end, for technical and business reasons, a horrible ISA won.

Not for technical reasons beyond `This compiled code I have won't run on
a different ISA', mind. It was mostly for the business reason: `I'm
going to go bust if I don't do what MS demands of me and only sell x86
MS-DOS boxes', 'cos that's the way it worked.

No no no.

But yes.

[He's behind you again!]

MS-DOS would let you sell MS-DOS installed on your computers by paying
them money. If you signed a contract saying you'd install only MS-DOS
and no other OS, you had to pay them less money.

That is how it worked.

Pentiums and Xeons were physically outperforming Alphas and
Itaniums.

But that is irrelevant because the PC war had been won long before then
- and the winning of that war by one side is why Alphas never had a
chance. Itanium - well, I dunno, you'll have to ask Intel what happened
to *** up that one. I have some ideas, but no inside info. It always
looked a bit crap to me, though; but I always liked the smell of Alpha.
Don't ask me why. Maybe I am some sort of pervert who likes to go
around sniffing design engineer's seats.

Urgh.

I know that was just a joke, but it makes me feel dirty just thinking
about it.

At far less cost. I already told you that the source code you
love to hate would, when compiled, run on Alpha and MIPS and Itanium as
well as x86. ISA was no longer a factor.

But it was and remains so. It matters because people do not want to
have to replace all their software - an expensive proposition - just
because they've bought a new computer. You're forgetting about the
customers in your enthusiasm for the technology.

Not many people know this, but the early work on 64-bit Windows was
done at Microsoft using Alphas, because Itanium was woefully behind
schedule, yet Microsoft was really worried that either or both of them
would eat the x86's lunch.

Why would MS care about that? All it would have to do would be rake in
the money selling lots of re-compiled software to people changing
hardware platform.

In the end, most 64-bit Vista ended up
targeted to Xeons because they won fair and square.

None of that has anything to do with the situation that I was talking
about. I don't know about the details of that particular battle so I'm
not going to comment much.

But it seems to me that the big problem that Itanium had was trying to
support x86 in hardware - that was bound to make it poorer. I assume
that Alpha was clobbered by lack of development funds due to DEC not
having the sales that Intel had at the time, and the reason for that was
the take-over of the PC ecosystem in an earlier phase by the MS/Intel
thing.

My argument for being resigned about a cruddy ISA is precisely the
opposite of what you assert. There are compilers that compile the same
code for different architectures so well that the ISA, elegant or
otherwise is completely irrelevant.

But that's not relevant: what matters is what people can do with what
they own - the things they have already paid for - now. People want to
run their existing software on their new computers. If they can't, then
buying a new computer is very expensive because they have to buy a new
set of software.

This is important - it's not just what the technology permits, it's
what's economically sane, given the currently installed business
software base of mostly commercial copyright controlled software. And
it's not economically sane, in the minds of most people, to just throw
away a big pile of valuable property in perfect condition just to
replace it with new property that serves exactly the same purpose. Or,
worse, new property that requires expensive staff retraining because
it's not identical (even if the new hardware and software is better all
round and would result in business efficiency improvements, many
managers would simply say `no' to spending the money on staff training
or software investment, not even considering if it might pay off).

That, including the training and maangement blinkers issues, is why
we're stuck with what we've got for now. For now, I tell you.

You need look no further than OS X to see how successfully software can
be made to work on utterly different ISA's. Or, if you must wallow in
the past, how well FAT binaries served an earlier generation of Macs as
they moved from 68000 to PPC.

Oh yes, but you've got to put the effort in. Did you see MS bothering?
Has MS ever attempted a job like that? No - it might well have compiled
an OS for a new CPU, but it never wrote the emulation layers needed so
that existing software could run without re-compilation.

(not that MS minded people selling non-x86 MS-DOS boxes, but you never
sold much like that; it was selling the computer without MS-DOS that MS
was against)

I think what it is is a cultural memory of those times - the times when
`MS-DOS' wasn't enough: it had to be `MS-DOS plus proper x86' and if you
didn't have x86, you were ***. I think that's what caused the
addiction to x86.
OK. I was there. I never saw much of that.

I was there. I heard complaints from people who had bought non-x86
MS-DOS boxes and then found that they couldn't run much MS-DOS software
'cos it was mostly compiled for x86.

Are you telling me that you found that people who didn't have x86 CPUs
in their MS-DOS boxes were happy with the lack of available software?

Nobody cares very much. Apple doesn't. Microsoft doesn't. It might be
ugly, but it is the fastest cheapest processor you can buy.

But it's not. If you look at supercomputers, you'll find that the
dedicated boxes all use `not x86' CPUs for best price/performance
figures. Okay, power consumption comes into it big time with
supercomputers - and other CPUs give you more `brains per watt' than you
get with typical x86 line designs.

You wanna check the latest top 500? 412 of the the top 500
supercomputers are EMT64T (Xeons) or AMD x86_64 plus a few IA-32. x86
boxes all.

Yep - clusters, all of 'em, yes?

There is one alpha, 2 Crays and 61 Powers, 2 NECs and one Sparc machine.
http://www.top500.org/stats/list/procfam
So you are just flat out wrong in the most spectacular fashion possible.

What I meant to express clearly was that I was talking about the
non-cluster-type of supercomputer. Don't tell me that the non-cluster
supercomputers are going x86? They're not, are they?

I'll admit I was taken aback by how fast the list has changed since
last I looked. 82% of the top 500 supercomputers are running on x86.
What a ghastly thought.

<shrug> They're not /that/ bad these days. Just so long as they've got
a decent OS, why worry? Ugly they might be when you look inside because
of the need to translate from the x86 external ISA to `what they really
do', but they don't half work quickly and for all that they use a lot of
power, they're cheap to buy.

The good news is that only 1.2% are running Windows.

I'm surprised it's that many.

OK, I care, but I don't count, and honestly it is an affectation, since
the instruction set of a machine does not matter a damn any more.

Phew! I'm getting maudlin. Lemme get another beer to cry into.

Yeah, you're letting your love of Alpha get the better of you. The
Alpha story is a sad one, but it's a story that happened after the old
PC ecosystem had been destroyed - that's why Alpha never got anywhere.

Well, I tried. Don't let the facts get in the way of a good argument.

ARGH!!!!

How many times do I have to explain that the events I'm talking about
did in fact happen before Alpha came out? I'm talking about what
happened before Win 3 was released!

PS. I never much liked Alpha. It had/has a really truly ugly ISA.

I know nothing about its ISA.

You
would have liked it, with that horrible load and store you find so
charming.

I only like that when it goes along with `bloody fast processing in a
very elegant and minimalist design at a low price'.

Otherwise it's just a pain.

At the end, it was far from RISC, with more opcodes than even
a VAX, which was always the poster child for CISC.
My heart stayed with the 11 and VAX, but Apha completely slew them for
price performance. It, in turn was completely slain by x86, on price
performance alone. It has nothing to do with any eco-system.

But I say that it has: Alpha lost on price/performance because by that
time, DEC was on the rocks, and Intel was the giant monster stalking the
Earth. DEC could not possibly deploy the development resources that
Intel had - and that was all because of the PC ecosystem being entirely
x86 by that time.

Alpha might have failed to sell for technical reasons such as
`price/performance' or whatever - but any technical failings it had by
comparison with the Intel competition are, I put it to you, a
consequence of the fact that Intel had more resources from huge sales
due to the fact that the PC world was `all x86' by then - and *that* had
happened due to events in the pre-Win 3 world.

See? I'm always right[1].

Rowland.

[1] Except when I'm wrong.

--
Remove the animal for email address: rowland.mcdonnell@xxxxxxxxxxxxxxx
Sorry - the spam got to me
http://www.mag-uk.org http://www.bmf.co.uk
UK biker? Join MAG and the BMF and stop the Eurocrats banning biking
.