Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!leah!bingvaxu!sunybcs!boulder!unicads!les
From: les@unicads.UUCP (Les Milash)
Newsgroups: comp.arch
Subject: Re:
Message-ID: <425@unicads.UUCP>
Date: 9 May 89 17:12:39 GMT
References: <2368@ogccse.ogc.edu> <1464@cfa.cfa.harvard.EDU> <141@dg.dg.com> <156@dg.dg.com> <658@pitstop.West.Sun.COM> <19088@winchester.mips.COM>
Reply-To: les@unicads.UUCP (Les Milash)
Organization: Unicad  Boulder, CO
Lines: 48

>In article <658@pitstop.West.Sun.COM> (Adrian Cockcroft) writes:
>there is also the issue of applications sw.
>"SPARC has over 500 applications"

i say "whoa!  wait a minute!" what does that mean? i thought that on unix,
you just typed "tar xvo ; make" and you had another application! :-) :-) :-) :-)

now i know it ain't usually quite that easy (-: boy do i :-) but
after reading all these discussions about how new processors mustn't break
old code, I'm kind of astonished that anybody could propose this as a
metric of anything except maybe "developers faith in the company" or
"cost of box after developers discount" or "faith that the architecture will
have faster next generations".

not meaning to flame anybody, really, i'm truly surprised to see this.

i mean if WhizBang Inc's new VRIW machine requires all short doubles to be
mersenne-prime aligned in memory or some ultra wierd thing, that might
shake off developers, might make it way too much of a hassle, and that'd be
an indication that the arcitecture was bad, or if it was just plain too
slow for the bucks cause of architectural reasons that would be too.
or if after the chip came out the only compiler i could get for 5 years
was COBOL++, that'd slow things down.  i also realize that if it's designed
into Everybody's Favorite UNIX Box (Suns ARE my fave right now) or even PC 
colone (-: a freudian typo--sorry :-) that indicates that there's alot of 
money that could be funding the next generation and so i could have some trust 
that there would be a next generation.

but it's not like MySWCo is gonna say "finally!  another register-window
machine, finally we have got another machine we can port to!"; ain't it
more like "hey they sold 20 million boxes to detroit last year, let's port
to that" and if it's unix with C then we do it.

excuse me if i missed what this thread was about, but i don't see how
"number of apps" has much to do with an architecture, except in the extreme
cases of extremely good or extremely bogus.  any success
SPARC has in this area seems to me to be a reflection of the fact that
Sun has it together (like they sold alot of 68Kboxes cause Sun [23]s were
cheap and good and the sw is pretty compatable across the line (admirably 
so, i'd say)), not SPARC arach per se; i bet they could have done just as well 
with an Rx000, or 29k or 88k or 386.  similarly, any success Prisma or Solborne
has is probably a reflection of how good Sun is as well (although any failure
they have they own).

Right?  I'm sure you'll correct me if I'm wrong....

(the perennially brash and hot headed) Les Milash
(maybe if i drank Lite Beer from Miller?)