[Simh] [SimH] Networking support

Sat Mar 12 15:31:32 EST 2016

Bill,

You probably need to date things a little and get a some perspective of
where a few of us are coming.   Just to set a few lines in the sand.  While
3Mb/s "xerox" ethernet has been around for about 5 years, the
DEC/Intel/Xerox Ethernet 10Mb/s spec was published Sep 30, 1980.   Per RFC
801, Arpanet was not officially schedule to switch from the old NCP to IP
until Jan 1, 1983 (although a number of folks like me had been working with
what would become IP/TCP for 3-4 years before that).  But "networking" as
we think of it today, had been around for over 10-15 years before that -
i.e. long before the rise of the PC.

Most large (mini/mainframe) manufacturer have their own networking.
 Consider DEC's VMS, the original IP/TCP implementation was written by me
and my co-workers at Tektronix at the time (1979-80).  DEC wants us to use
DECnet.   We also have CDC, UNIX boxes and even a IBM mainframe.   We
needed something that could span the OSses.  It was easier for us to write
our own implementation because what we wanted was a network that spanned
our companies use, not one manufacturer.

Moreover "Metcalfe's Law" is also very important (the value of a network is
proportional to the exponent of the number of the things connected to it).
So ... each manufacturer had their own scheme - both HW and SW.   The
widest "general" system was the Arpanet "IMP" system that DoD paid to have
interfaced to a number of different systems; although as has been discussed
PDP-10's, 11s and eventually Vaxen were a large number of the systems on
that network.  So in fact, that was the most valuable network because it
had the most systems connected to it that had cool things that could
used/shared.

Actually, the largest reaching network at the time was the Unix UUCP
network, which allowed email and file transfer (al biet a tad slowly).
There were thousands of sites on it.  In fact the UUCP site "ccc" pre-dates
ccc.com being registered to me because in those days people could not
register a domain name.  But "usenet" could not do things like telnet and
was a good bit less formal that the ARPAnet or the Internet.

Roll forward and think of PC's.   Please remember with the PC's
pre-ethernet  too were often "networked" most often with ARC-net HW (75 ohm
coax - very cheap), running either Netware or LAN-Manager.   Like, DECNet
and the systems from the mini's and mainframes, it allowed PC to send files
around, share printers, send email etc.   But like all other proprietary
systems, it was closed so it's not talk to your Vaxen or UNIX systems etc.
The size of these networks tended to be the size of your office during
Word/Excel et al.   Which was fine for a lot of people... but....   it's a
little like having a local walkie-talkie that is not connected to the
larger network and you can talk to a few people.  When you finally realize
you need to talk to someone outside your firm, you note how limited it it.

So... between the US Gov paying for the SW to be written to support a # of
OS's they cared about, the rise of UNIX, and MetCalfe's Law, IP became the
protocol and ethernet became the HW that "stuck" - i.e. the rise of the
Internet.

Now consider that by the time the PC and BBS system that PC's used come on
the scene you have a quite a different view.  Also, BBS's were really not a
network in the same way the mini's and mainframes worked.  A person that
knew a phone number and had an account to call a remote system.   But this
is quite different from how the DECNet, SNA, Arpanet, etc (or for that
matter UUCP) worked.

Back to your question...  if we want a simulation system to support
networking in the way most of us think of it, we need two things.   First
the HW needs to emulate some know HW that was developed and released by the
manufacturer.  Second, you need the OS support for same.

Frankly, it is probably not worth investing a lot of effort into writing
the HW emulation unless we have the SW to drive it.   And frankly, you need
to think how you will use it.  Modulo Johnny and the cool folks running
HECnet (a large world wide network running DECnet over the Internet), you
probably will want to have Internet functionality to be able to access the
systems.

The good news is that a number of folks developed implementations for
almost most of the major OS implementations and many of the manufacturers
eventually picked them up (DEC would eventually take the Tek/CMU IP
implementation in house).

The bad news is I fear except for a few cases where the manufacturer picked
it up and made a product, some of those stacks (like the HP-1000 and
HP-3000 stacks from BBN) have been lost.   If someone has those it would be
cool, but I have not seen those bits since the early 1980s.

Clem

On Sat, Mar 12, 2016 at 2:36 PM, Bill Cunningham <billcun at suddenlink.net>
wrote:

> What I meant was that I remember on early PCs using an rs232-c line for
> using the old BBSes and compuserve before it was an ISP. 10 cents a minute.
> I had several modems 300, 1200 and 2400 baud modems.
>
>     These even older machines may have had hookups within a company. Even
> one building connecting 5 or so machines. Serial would've worked fine. And
> was what was used. I was thinking with maybe 4-5 PDP8s a company would use
> some kind of networking. Perhaps not back then. I was only aware of pdp11
> and vax being "network possible". I guess I was wrong.
>
> Bill
>
>
> ----- Original Message -----
> *From:* Clem Cole <clemc at ccc.com>
> *To:* Anders Magnusson <ragge at ludd.ltu.se>
> *Cc:* SIMH <simh at trailing-edge.com> ; Bob Supnik <bob at supnik.org>
> *Sent:* Saturday, March 12, 2016 2:24 PM
> *Subject:* Re: [Simh] [SimH] Networking support
>
>
> On Sat, Mar 12, 2016 at 12:25 PM, Anders Magnusson <ragge at ludd.ltu.se>
> wrote:
>
>> DG-UX or MV/UX?
>>
> Which was the rewrite of System V ?? i.e. System V cmd system, but
> internally developed System V SMP kernel -- I want to say DG-UX maybe; but
> I'd been a long time and many beers ago - I've forgotten the name.   I
> remember it was a very clean UNIX implementation.   Nice locking structure,
> easy to debug, etc...
>
> Locus was working on different projects with Ultrix, Tru64, VMS, AIX,
> SunOS, Solaris, HP-UX, Apollo, DG's UX, some work for Pr1me, ISC's 386/ix,
> Intel's 386 port, SVR4 for the AT&T/UI guys, and Intel's Paragon at the
> same time.  At one point, I had the OS release schedules for HP, DEC and
> Sun all pasted on the wall behind my desk.  I used to say LCC got to see
> everyone's dirty laundry in those days.  As I said, I do remember the DG
> Unix re-implementation was very easy to work on (I will not say which one
> we cursed the most).
>
>
>
>
>> The DG ethernet card has a 82586 on board.
>>
> As I said, many beers ago. I'm undoubtedly mixed up a couple of the
> systems, since we had so many we worked with in those days.  I remember the
> AMD chip was a lot easier to program than the Intel device. That said, I
> suspect that I have the docs on the Intel chips somewhere, but it sounds
> like others have the DG docs which are going to be better for simh purposes.
>
> Clem
>
>
> ------------------------------
>
> _______________________________________________
> Simh mailing list
> Simh at trailing-edge.com
> http://mailman.trailing-edge.com/mailman/listinfo/simh
>
>
> _______________________________________________
> Simh mailing list
> Simh at trailing-edge.com
> http://mailman.trailing-edge.com/mailman/listinfo/simh
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20160312/2c02d445/attachment-0001.html>