Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 (Tek) 9/26/83; site orca.UUCP
Path: utzoo!linus!decvax!harpo!ihnp4!zehntel!tektronix!orca!brucec
From: brucec@orca.UUCP (Bruce Cohen)
Newsgroups: net.micro
Subject: Re: Who's Next?
Message-ID: <452@orca.UUCP>
Date: Fri, 6-Jan-84 14:12:28 EST
Article-I.D.: orca.452
Posted: Fri Jan  6 14:12:28 1984
Date-Received: Sun, 8-Jan-84 01:14:55 EST
References: <227@dsd.UUCP>
Organization: Tektronix, Wilsonville OR.
Lines: 59

--------------------------
Mike Spann says:
>>  
>>    No, I don't believe that UNIX(tm) is the next popular O.S. for
>>  micros.  If it weren't for the public abhoration for anything INTEL
>>  does until some outside vendor changes the name and sells it, I would
>>  say RMX-86 would stand a good chance.  It has reasonable multi-tasking
>>  facilities and a semi-acceptable command interpreter.

It does my heart good to hear someone outside of Intel (besides me) say that.
While I don't agree that the command interpreter is acceptable, most
everything else about the system is quite good, from the cleanness of the
code to the performance.  The file system is *better* than the Unix file
system in a large number of ways including dynamically configurable device
drivers, several different flavors of file system (physical, flat numbered,
and hierarchical named) available on a per-device basis, and a cleaner device
driver interface.  Bob Beck, take a bow.

I have written system code and user code for both RMX-86 and RMX-88, its
little brother, and I find that the ability to use many small tasks to do
a job, rather than a few monolithic processes as in Unoid systems, is a
big win.  The inter-task communication is quite straightforward to use, and 
moreover performs well.  The fact that the kernel is actually a small piece
of the OS means that the I/O system could be built using multiple task and
inter-task communication.  This allowed me to design, write, and debug a
terminal/tty driver for RMX-88 in 8 weeks, and port it to RMX-86 in one week,
without ever having written a device driver before.  It exceeded the
performance goals without any tuning at all.

>>  Little
>>  effort has been given to taking one of the multi-tasking packages and
>>  build reasonable support utilities over it.  (You pick it, VRTX or
>>  MTOS are supported on several machines including the 68k and '86).
>>  The only integrated package for a micro that I know that does all this
>>  is RMX-86, and it is an INTEL product.  It also only runs on an 8086
>>  and (here goes again) is written in PLM-86.

So true, *SIGH*.  And  a source code license for RMX costs ~$50k, and you'd
still spend 3-5 man-years porting the parts that are in PLM to another HLL
like C (unfortunately a lot of the kernel is in ASM-86).  So it's not a
garage shop project.  If it were, I could get a team to do it tomorrow.  Worse
still would be duplicating the work.  Even considering that a lot of the time
and money went into false starts and internal boondoggles (I was there, so I
can speak  with some authority on the subject), the development of RMX-86
cost many millions of dollars, and tens of man-years.  Even done right from
the start, it's not the sort of product you would want for the initial
project of a startup company, with the capitalists breathing down your neck
for return on investment.

Still, it would be possible for some enterprising souls to get a product out
for the IBM PC class machines and get some revenue while generalizing the
system for porting to the 68K, 16K, and whatall.  Anyone out there who wants
to discuss that as project, (as a profit-making venture, of course) send me
mail.

				Bruce Cohen
				UUCP:	...!teklabs!tekecs!brucec
				CSNET:	tekecs!brucec@tektronix
				ARPA:	tekecs!brucec.tektronix@rand-relay