Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!ucsd!ucbvax!CAEN.ENGIN.UMICH.EDU!pha
From: pha@CAEN.ENGIN.UMICH.EDU (Paul H. Anderson)
Newsgroups: comp.sys.apollo
Subject: Re: ADUS conference news
Message-ID: <4d7dc3d09.0017b5e@caen.engin.umich.edu>
Date: 19 Oct 90 13:31:09 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 62


	
	In article <9010151904.aa17376@concour.cs.concordia.ca>, goldfish@CONCOUR.CS.CONCORDIA.CA (-- Paul Goldsmith) writes:
	  Before you get out your hankies and start crying over HP's costs for
	  putting OSF onto an Apollo, the Apollo DN4500 was demonstrated running
	  MACH (which is the OSF kernel) at the Montreal Computer show. (I
	  didn't see it myself, however another analyst saw it there.) Once the
	  kernel runs, any other excuses for not putting the OS up are STRICTLY
	  MARKETING.
	 
	I worked on the original port of Mach to Apollo hardware back in 1988.
	Getting the basics (compiling and booting) is pretty easy.  But the fact is
	that "once the kernel runs" you've just barely started.  All the support
	costs are directly proportional to the number of different architectures you
	want to support.  And the number of different architectures is amazingly
	high.  Did you know that the dn3000 comes with two different MMUs?  (One of
	them is actually the dn3010 if you want to get picky).  Every time a new
	release comes out, Apollo has to beta test it on all the different
	architectures.  Add in the different display, memory, disk, and net options
	and the number of permutations is unbelievable.
	 
	Marketing certainly goes into the equation, but there are in fact very real,
	and very high, costs associated with supporting (as opposed to just
	releasing) OSF on all the current Apollo hardware types.
	 
	I doubt that HP is planning to strand owners of the current products.  By
	the time HP drops support for Domain/OS, no one will be using Otters
	(dn[34]xxx) any more, for the same reason that no one uses Terns (dn[46]60)
	any more.  It just won't be cost effective to keep them operating.
	 
	Of course, it would be really cool if somebody made Otter OSF available as
	an unsupported product.  I doubt that we'll see this.
	
Apollo has demonstrated OSF/1 on DN4500's at trade shows.  The HP contributed
port to OSF includes support for the DN2500/DN3010/DN3500/DN4000 and DN4500.
Not to say that it is done and tested for all platforms, or anything, but an
awful lot of the work has already been done.

Furthermore, with pooled resources developing the kernel, the theory behind
OSF is that individual manufacturers won't have to re-invent the wheel simply
to come up with a robust, usable system (as Apollo did with Aegis - for good
reason at the time).  Therefore, the total cost of offering the system on
a given platform will be substantially cheaper than offering Domain/OS on
a new platform, since the machine specific portion is smaller and better
defined.

Reasonably high level marketing people told us earlier this year that OSF
would be supported on DNxxxx series machines.  The announcement at ADUS
indicates that they were lying to us.  Independently of whether or not
this decision is good, customers don't like being lied to.

Yes, it costs money to test different variations of SW and HW, but that is
what we buy the machines for in the first place!  If HP/Apollo can't support
what it sells, then I don't feel too badly about it losing the market share
it picked up when HP bought Apollo.

Paul Anderson

Disclaimer - If this is a flame, it is, as always, directed against
upper level management at HP/Apollo.  For some damn-fool reason, I
thought it would be an improvement over the old Apollo upper level
management.  I guess I was wrong.