Info wanted-- IBM 3274

From: Ethan Dicks <erd_at_infinet.com>
Date: Fri Jul 24 10:14:43 1998

>
> In a message dated 98-07-24 07:59:41 EDT, you write:
> I guess the 3274 could be called a computer here. it will process information
> and if it can do something, it will take care of simple local tasks itself
> rather than bothering the mainframe with things the 3274 could do itself.

Practically speaking, though, it's as much a computer as an Adaptec ISA
SCSI board. True, both do have firmware and to have some sort of
processor, technically making them a computer, the computer part is so
embedded as to vanish from sight.

> The 3274 is called a "terminal controller"...

And is a PU Type 2 (Physical Unit) in SNA parlance (as opposed to a Front
End Processor (FEP), a PU Type 4, or a CPU itself, a PU Type 5). All
have their own little dance to do in the grand scheme of things on an
SNA network.

> ...the 3278 and its relatives use the
> IBM SNA protocol over 93 ohm co-ax, the 3274 can AFAIK connect over an
> ordinary serial line.

First of all, only the voltages are ordinary (-/+12). It's going to be
SYNCHRONOUS serial, meaning that the data clocks are on pins 15 and 17,
with NO start bits, NO parity bits, NO stop bits, just 8-bit bytes, one
after the other. SYNC comms get better throughput than ASYNC for this
reason (8-bits vs 10-bits per byte). The Z8530 SCC chip can be made
to do SDLC SYNC I/O. We used it on our SNA products.

> ...In which case all you need to do is write drivers
> for whatever machine you choose as the host...

I used to make and sell HASP/3780/SNA boards for VAXen (I still have a
bunch). I am here to tell you that while it _can_ be done, it's a major,
major undertaking. Making 68000 boards that simulated a PU Type 2 (3274)
was a significant commercial endeavor, 10 years ago. I worked on the guts
of the code, on the application end, the device driver, all of it. I
would _never_ consider re-implementing SNA as a non-paying project. In
effect, your "drivers" would have to emulate a 37x5 FEP sufficiently enough
to fool real Blue hardware.

We had these HP line analyzers, I forget the number, with a 40-column
screen and TU-58-sized tape drive. In addition to tracing traffic, you
could write programs (subject to severe memory limitations) that could listen
for some stuff on the line, respond appropriately and simulate portions of
the data stream. We had one program that took two of us to write and debug
that could lead a PU Type 4 (ours and a portable terminal called an Informer)
through the login process, and that was about it. A connection and one data
packet. With a better protocol engine (to properly keep track of RR numbers),
it might have been possible to do more, but we maxed out the HP box.

Let's put it this way, I'd put writing a driver (assuming you had a compatible
SYNC serial port) to talk to a real 3274 and then to terminals, at least as
complex as writing an implementation of PPP from scratch, possibly with
TCP/IP thrown in on top, just to make it a little harder. SNA is extremely
non-trivial. Just because it's classic, don't make it simple.

OTOH, the 3274 is a nice, *portable* piece of commonly used Blue Iron that
would make a fine addition to any collection. _Using_ it is a different
matter entirely.

-ethan
Received on Fri Jul 24 1998 - 10:14:43 BST