At 12:00 AM 12/27/95 -0500, Automatic digest processor wrote:
>Date:    Tue, 26 Dec 1995 11:08:36 MST
>From:    "F. Alfredo Rego" <[log in to unmask]>
 <snip>
>Speaking about the Internet way, I have a little surprise for all of you
>coming up in a few days.  The surprise is about 80% technical and about
>20% political, regarding these very issues, with some connections to
>ideas that will shake some people quite hard.
 <snip>
>In a nutshell:  My vision for the HP3000 and IMAGE is of something simple
>that works reliably and economically.  HP has a golden opportunity to
>be the only providers of such low-key, non-hyped-up BUT ESSENTIAL resources.
 
Alfredo,
 
You say:  "non-hyped-up BUT ESSENTIAL resources."  Who is to say what is
essential?  In our environment, Internet services are "essential".
 
No, haven't been talking to Wirt lately.  Maybe that we are both from
educational system backgrounds might make us sound similar in our opinions.
 
We use an HP 3000 in highly "cutting edge" and sometimes "bleeding edge"
environment.  Internet software and techniques are more the norm than the
exception here.  We need solutions years ahead than the product cycles found
on the 3000.  For instance, Internet solutions that are just starting to
make HP 3000 inroads today we needed 4 years ago.
 
In 1992 when our campus network sprang up and interconnected buildings all
over campus, our HP 3000 stood alone like a castle that had a network moat
around it.  Interconnectivity came in the form expensive, clunky
drawbridges, like HP's TAC cards that had poor performance to cost ratios.
Yes, there were solutions to connect, but they did not follow Internet
models.  The availability of host Internet servers on an HP 3000 platform
was near to nothing.  And let us not forget about the non-standard Internet
terminal connectivity issues today.
 
With the release of MPE 4.0 (1993), allowed direct IP connections from the
campus network.  We experimented with a few prototypes and we implemented
host based telnet and gopher servers that year on the HP 3000 (a year before
NS/open!).  Our telnet server was also a terminal translator - it converted
absolute HP term cursor addressing (highlites) to Internet standard VT100
cursor addressing.  This allowed us to provide host based telnet to a large
population of on-campus users, using the SAME program that was running under
WRQ reflections on serial connections.  I hacked together a standalone
version of that that program and is available via ftp as NQtelnet located on
opus.
 
Our HP 3000 gopher server was in production late 1993.  Unlike typical
gopher servers, this service is authenticated - information like your
grades, schedules, charges at the bookstore etc. are presented in
_personalized_ menus (a replacement of the host logon concept - many people
still don't understand this).
 
Gopher Stateless protocol allowed thousands of campus end points to connect
and use this service.  And nobody noticed any loading problems on the host
3000.  Instead of believing the myths that you must avoid DBOPENS at all
cost, I setup a test and found that you should avoid the FIRST dbopen at all
cost.  Second and third dbopens don't cost much.  So, I keep the databases
open with the monitor program and implemented a database access service no
one believed should work on the scale it does today.
 
So, we grew our own Internet solutions on the 3000 (I'll be glad to get out
of the Internet software business when the HP catches up!).  Now we are
looking at DCE.  I'm tickled pink that the 3000 has the DCE core services
(Posix allows it), but it is DCE 1.0 based and no file integration (DFS).
Again, the 3000 isn't playing with a full Internet deck.
 
As Jeff Kell pointed out, the Internet weakness of the HP 3000 can also be
an asset - sometimes (but I rather manage the weakness and have the
service).  And you're right Jeff, probably good reasons not to emulate MPE
at all - no matter what system.  And Alfredo's right, Posix on the 3000 will
be the best compromise between MPE and Unix - might be even better.
 
The "death of Unix" Ron is pointing out is nothing more than a re-birth of
better OS's w/Posix standards.  Why?  The DCE is coded around Posix
standards and to fully support the DCE you must reengineer all those flavors
of Unix (like BSD unix) -
 
If I follow Ron's argument, are we back to saying that the 3000 is so great
because we don't reengineer?
 
Keep on eye on the DCE.  I hope that is what Alfredo's "surprise" is all
about.  Using the DCE RPC and IMAGE as object's.
 
Eric Schubert