Xref: utzoo comp.arch:6582 alt.next:83
Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!ukma!gatech!fabscal!dorn
From: dorn@fabscal.UUCP (Alan Dorn Hetzel)
Newsgroups: comp.arch,alt.next
Subject: Re: The NeXT Problem
Keywords: NeXT
Message-ID: <521@fabscal.UUCP>
Date: 15 Oct 88 16:18:19 GMT
References: <26435@ucbvax.BERKELEY.EDU>
Reply-To: dorn@fabscal.UUCP (Alan Dorn Hetzel, Jr.)
Organization: Fabscal Systems, Inc.
Lines: 157

In article <26435@ucbvax.BERKELEY.EDU> pchris@ucbarpa.Berkeley.EDU (Chris Perleberg) writes:
>
>It seems that the NeXT machine may have a few problems:
>
>1) Outdated Processor Technology: NeXT just missed the wave of fast RISC 
>   processors.  The 5 MIPS 68030 is completely out performed by the currently
>   available RISC chips (Motorola, MIPS, Sparc) that run at approximately
>   20 VAX (they claim) MIPS.  In a year or two, ECL versions of some of these
>   RISC chips will be running at 40 to 50 MIPS.
>

On this point, I'm not yet convinced that RISC machines are the only way to
go.  They will have their uses, but don't expect to see the death of CISC
processors anytime soon.  RISC places some significant additional strain on
memory bandwidth for instruction fetch, and while in some programs this can
be offset by instruction caches, in other programs it cannot...  Also,
Motorola and the other purveyors of 68xxx family products can be expected
to keep improving their line as well (68040, higher clock rates, etc.)
There is also the fact that the NeXT is a passive backplane system, with the
CPU on a removable, (and therefore replaceable) board.  Mach also supports
multiple processors.  While not exactly related, the system also contains
a 56001 DSP processor, which will aid GREATLY in some types of processing
which micros haven't been much good at previously.

>2) Non-Standard NuBus Implementation: A small company like NeXT can't hope to
>   create a competitive 3rd party board market for a non-standard bus.
>

"a small company like NeXT"   ...   backed by H. Ross Perot.

(Also, a small company like Apple has started a couple of busses itself)

>3) Non-Standard Software: What software company would develop software for
>   the special features of just one computer (NeXT Step)?  How many copies of 
>   this software can they possibly sell?
>

Well, Mach is unix like enough to make portation of significant quantities 
of unix software not too onerous a task.  As far as NeXT Step, I imagine it
will do OK, considering the fact that IBM (yes, Big Blue) has licensed it
to use on some of their systems as a user interface.  I figure they might
sell a FEW copies.

>4) Slow Optical Drive: In the past, optical drives have been significantly
>   slower (seek times) than magnetic drives.  What is the advantage of the
>   optical drive?  Cost must be less than that of the larger 330Mbyte $2K
>   magnetic drive.  But NeXT will be hurt once benchmarks come out for its
>   i/o performance (using the optical drive).
>

a) It may not be as slow as you think, electro-optical drives are near the
   leading edge of the optical pack in speed.
b) one (1) advantage of the optical drive is that it is a 256Mb  REMOVABLE
   disk subsystem.  More blank disks cost about $50 each.
c) Yes, costs will be MUCH less than the 330Mb magnetic disk, which still
   takes multiple platters and heads;  The optical disk uses one platter.
b) Even assuming their benchmarks will be dismal (and I don't think they
   will), purchase decisions are not made by benchmarks alone.

>5) Software Not Ready: The 9 month delay (optimistically) until solid software
>   exists could kill NeXT, as Sun & Apple prepare competitive systems.  Sun
>   will probably keep to open systems and set some new standards, while Apple
>   will probably stay proprietary.
>

Late software hurts, but they appear to have waited long enough with the
introduction of the system itself that their software timetable can be met.

>6) Sun (I have heard) has sold 15,000 workstations to universitys.  How many
>   can NeXT expect to sell with its slow processor, non-standard bus/software, 
>   slow drive, and late software?
>
>

We will have to wait and see, but considering the fact that NeXT is going to
concentrate exclusively on universities as a market, they may do pretty well.
(A footnote, one local university here has already ordered 100 of them)

>Solutions to the above problems: What NeXT should do.
>
>
>1) Develop a RISC based NeXT implementation as soon as possible.  The advantage
>   of Unix (Mach) is its idea of source level portability, rather than binary 
>   level compatibility.
>

Oh, pish, you have RISC on the brain.  A second (or third) processor board
can always be added if warranted.  You were talking about lack of software,
well, there is a lot less software for RISC than for the 68xxx family.  You
complain the software will be late, but here you want to make it later..!

>2) Make "stub" boards that convert standard NuBus boards to the NeXT version of
>   NuBus.  These "stubs" would be placed between the NeXT slots and the
>   standard NuBus boards.  Longer Term Goal: Make the NeXT NuBus an
>   international standard, much as Apple made its version of NuBus a standard.
>   Possible Solution: Change the bus NOW to a standard, provide board
>   converters for computers with the current bus.
>

I'm not sure there is any difference between their bus and the NuBus except
clock rate.  If that's the case, the 68000 is pretty tolerant of slower than
expected devices, so maybe slower boards will work.  If not, consider that
IBM got away with introducing another bus, and now there are cards for it.
Also, given the SCSI port for disks, the wide array of goodies already on
board, and the substantial amount of memory in the base system, many users
may be able to "limp" along with the basic system for quite some time.

>3) Make NeXT proprietary software into standards, and beat Sun at its own game
>   before Sun beats NeXT.  This may mean making NeXT Step an overlay on X
>   windows.  The important thing is to develop a standard that can be (and is)
>   used by all computers.
>

Given the nature of the display hardware, X-Windows won't be any big deal to
port.  Also, display postscript (even though i'm not that fond of it myself),
has good potential for becoming a display standard interface (at the lower
level).  And at the risk of becoming repetitive, NeXT Step will get a big
boost towards becoming a standard by IBM's licensing of it.

>4) Give users the option of scrapping the optical drive and replacing it with
>   the 330 Mbyte winchester.
>

The optical disk drive is, among other things, the software distribution
media for the system (no more swapping a zillion floppies to install your
favorite LARGE software package).  Economies of scale attained by placing
one as the "floppy" of each system make it quite inexpensive.  I suspect
many users will want two of them.  Their winchester prices are a little on
the high side (hell, quite a bit), but given the SCSI interface, I see
strong third party availability of mass storage and tape backup and such.

>5) Work like hell.  Adopt standard software where ever possible.  Distribute
>   work to 3rd parties.  Standardize where ever possible to beat Sun at its
>   own game and to make the users happy.  Allow Apple to be proprietary and
>   dig its own grave.
>

Given the nature of workaholics like Jobs and Perot, I expect they have been
working like hell and then some since they started this thing.  The deal with
IBM demonstrates their intent to standardize.  As far as universities go,
Mach is a pretty reasonable choice for an OpSys standard, given its multiple
processor support and ongoing development at CMU.

>6) Standardize and Speed-Up -> Increase sales.
>
>

Hey, if they could control the sales numbers, they would have it made for
sure.  Since they can only sell so many at a time and still support them
well (they have to grow), I applaud their decision to go for the university
market first.


(Stepping down of my soapbox, I remain)

Dorn
gatech.edu!fabscal!dorn