Enter your search terms Submit search form
	Web pdp-10.trailing-edge.com

I felt like a need to say this for some reason.  I'm not sure if it's
a generally recognized thing here.  It's about software modifications
and the computing environment we'll be moving into soon.  Some things
on the tcp-ip list and discussions about hiring another systems
programmer got me thinking about it.

The point of this is that we really must have source code to as
much stuff as possible when moving to Unix, and that we really
do need programming done.  We can't just use what we get from
the vendors.

Consider what would happen if we didn't have Monitor source code on
the NIC.  We'd probably be crashing every hour.  Getting DEC to fix
anything would take months if ever.  Also think about all the
software on the NIC.  I can hardly think of ANY program we have
that doesn't have some NIC bug fix or change in it, even if it's
only a line or two.  In the counterexample that comes to mind,
SCRIBE, we probably would have NIC stuff in it if we could get the
source code.  I think the same goes for almost any program we have.
I mean, we even went so far as rewriting our own compiler.  

You might think that this is only because we are using an old unsupported
operating system, but that really has nothing to do with it.
Thinking that moving to Unix will eliminate these NIC changes is
completely false.  People tend to think that since Unix is the
new thing and since everyone's using it, that it must work real
well and be properly supported.  That's not true at all, and just
playing with a SUN for a few hours will convince you of that.
Most vendors just give you whatever comes with the standard Berkeley
Unix release with a few mods for their machines.  SUN uses the
standard print spooling and mail system software that comes with
Unix, and it's just plain awful.  I mean, if a huge vendor like
SUN still distributes crufty old software, I don't see how we would
ever get along without a bunch of programming.

So, as far as Unix is concerned, we are going to have a lot of work
to do.  From what I've seen, programs in Unix are less developed
and less sophisticated than what we've been using for years.  It's really
a step backwards as far as software goes.  (Well, in general at least.
There are SOME nice things in Unix).  And I'm not trying to knock Unix
(that much), just trying to point out that we are going to need at least
as much and probably more software development to use it than we have
on the 20.

It should be noted that there will be lots of programming to do
as far as systems stuff goes, and that it will probably take some
time before it gets to the point of being a nice comfortable
environment.  I think it'll take at least of year of programming
before we get to the point where most of the things are working
the way we want them to be.

I'd also like to dispell some myths people might have about switching
to this 'new wonderful networked workstation environment'.  Basically,
it really isn't much of an improvement over what we've got now.  You
do gain a few things, like guaranteed CPU time on your own machine,
and a bit mapped display, (and Unix, if you think its an advantage).
BUT, you don't gain reliability or independence from centralized
services.  If a file server goes down, everyone loses.  If the network
jams or breaks, everyone loses.  If the network or file server becomes
overloaded, EVERYTHING slows down.  If a gateway dies, NOONE can talk
to the net.  It is still a very tightly coupled system, and any one
breakdown can affect everything.  With lots of extra equipment (and
money) it could be designed to be more reliable and independent, but
that's not within our budget.  (And in fact, a properly designed
workstation environment would cost much more than an equivalent
mainframe environment).  The point is that the workstation environment
is very much like the mainframe/terminal environment.  Instead of
everyone relying on one machine, everyone relies on a whole bunch of
distributed machines.  Its a much more complex system and there are
more things that can go wrong.  I don't think you should expect that
it will be much of an advance over what you are using now, but we'll
try to make the best of it.

That's it.