[mod.computers.workstations] Information wanted about Apollo

giebelhaus@umn-cs@hi-csc.UUCP (02/11/87)

The Ring Network:
  I find the DOMAIN Ring network is much better than ethernet (if you
set it up like the site planner's guide suggests).  I don't really
see the difference between installing a DOMAIN network outlet and a
tranceiver; especially if you don't use vampire taps.  I think the
outlets are easier to install than BNC connectors.
  Sure it is a ring so you have to run the wire there and back, but
the wire is much cheaper and for us, running two wires costs the same
in labor as running one wire.
  For debugging and the such, a good ring network is a sure win.
Stallings has a good discussion of ring vs. baseband in his book
"Local Area Networks".  In short, with the ring, errors are quickly
pointed out by surrounding nodes and the ring is better at resource
allocation.

Apollos in academic circles:
  I'm not sure about the academic circles.  I'm in Minnesota so let
me use the U of Minn as an example.  The U of M has both Apollos and
Suns.  Most of the Apollos are used by professors and grad students.
The undergrads for the most part use the Suns.  It may appear at
first glance that there are no Apollos at all at the U.  In reality
they are just hidden from the undergrads.  I don't know if this means
that the Apollo is better so is being saved for the higher priority
users or what.

Apollo and UNIX:
  The source is somewhat available.  I have the UNIX source right
down to the kernal and I didn't have to pay a dime for it.  You do
have to have a UNIX source license, sign a non-disclosure agreement,
and show some need.  When it comes to UNIX, showing the need isn't
difficult.  The rest is a matter of paperwork.  Yes, I believe NFS is
now ready for ordering.  I'm not sure what delivery is on the thing.
I'm still waiting for RFS, though.  Something with file locking and
the such built right in rather than added as after thoughts.
  The performance of the UNIX is pretty good.  A few commands do run
slower, but as is mentioned, this is being fixed.  I almost wish that
it were not because it is being fixed at the cost of some robustness.
Aegis is much more robust than UNIX.  If you look at benchmarks
published by publications such as Dataquest you will see that the
Apollo 3000 runs faster than the Sun 3/50 series.
  Porting is pretty easy also.  Of course programs that use /dev/kmem
have a problem, but I don't believe in using kmem anyway.  I think it
is very poor software engineering to steal variables.  What happens
if someone changes a data structue in the operating system?  Not very
modular, is it?  There are a couple other problems, but they are very
small and don't require major hacking at all.  The two largest
programs I have ported are news and c-kermit.  No problem at all with
either of them.

  Over all I am very pleased with Apollo.  They have much more than
just UNIX.  Programs such as netmain which analize networks, usage,
and hardware so that problems can be found before they are critical.
For example, I have never lost a disk as I am warned when a disk
starts going bad so I replace it befor it fails.  One should be
warned that I am not a unix hacker and I don't like to hack.  I like
things a little more stuctured and modular.