Re: OT: BBC site's old computers quiz

In article <timstreater-38A88F.12032821102008@xxxxxxxxxxxxxxxxxxx>, Tim
Streater <timstreater@xxxxxxxxxxxx> wrote:

In article
<1ip5d4r.1j1jvmqu9psvsN%real-address-in-sig@xxxxxxxxxxxxxxx>,
real-address-in-sig@xxxxxxxxxxxxxxx (Rowland McDonnell) wrote:

Tim Streater <timstreater@xxxxxxxxxxxx> wrote:

real-address-in-sig@xxxxxxxxxxxxxxx (Rowland McDonnell) wrote:

Tim Streater <timstreater@xxxxxxxxxxxx> wrote:

real-address-in-sig@xxxxxxxxxxxxxxx (Rowland McDonnell) wrote:

Tim Streater <timstreater@xxxxxxxxxxxx> wrote:

real-address-in-sig@xxxxxxxxxxxxxxx (Rowland McDonnell) wrote:

Jon B <black.hole@xxxxxxxxxxxxxxx> wrote:

[snip]

It's part of the
internet generation,

You think people weren't using the internet 25 years ago?

People were, but not that many. We'd only just started to install
ethernet, back then (at SLAC, this was).

What does Ethernet use at SLAC have to do with the number of people
using the Internet?

Only to the extent that not many people were networked, even at Labs.

You mean labs; but I sort of get your point.

Thing is, Ethernet is merely one of many networking protocols. `Just
starting to install Ethernet' doesn't mean that the place wasn't
networked before, y'see.

Well, in practice it does.

But computer networking began in the 1950s.

There were, so I thought, plenty of things like this around prior to the
1980s:

<http://en.wikipedia.org/wiki/Cambridge_Ring>

Sure, but not having a general overview of networking in the 70s I don't
know whether the word "plenty" is fair or not. At CERN at that time we
were connecting data acquisition computers to a larger central host at 1
and 5Mbps, but it was point to point again. As with your ref above we
had to make our own hardware and software, but it did get developed to
the point where remote editing, job submission, etc were possible, and
where it was possible to boot your PDP-11 remotely (the "firmware" to
support this on the PDP-11 was 8 bytes long).

So it became possible to entirely dispense with cards and paper tape.
Not everyone trusted this, though. I remember one guy who insisted on
keeping his 24,000 punched cards and used to have 12 trays of them
stacked on a tea trolley that he rolled around.

I was in a position to get some kind of overview of network hardware
and software from the late 60's on. Or at least what happened outside
the experimental networks in academia. (In 1969 I was working as a
comms guy for Reuters then moved to DEC's special systems group between
1973 and 1976, after which I started my own systems company and
continued working mainly in comms related areas.)
So here's how I saw it (mostly from memory, some of it with parity
errors)

Computer to computer communications was very rare in 1970. The nearest
to it were IBM's remote batch systems, which were card reader/punch
stations on a long stick. IBM bisync was de facto protocol. A ghastly
half duplex, optionally polled/multi-dropped disaster. It also
supported block mode terminals, the dreaded 2780. Everything was point
to point. The cost of long leased lines was terrifyingly high.

There were a number of hardware solutions outside IBM. Gandalf springs
to mind. The British bit of STC had a papertape reader punch thing in a
beautiful teak cabinet that communicated with like minded furniture at
1200 bits/sec over dial up. In 1970, I wrote an emulation of it for a
PDP-8 and some special hardware we built. We typeset and transmitted
stock market price tables for newspapers round Australia with it for
many years. Reuters had a joint venture with the delightfully named
Ultronic Systems Corporation for collecting and disseminating real time
share prices to brokers and traders on special 'desk units'. It was all
bespoke hardware, no software at all. 1200 bits/sec leased lines
multidropped profusely to save interstate line rental. In 1969, the
Australian Post Office bragged they had 100 modems in operation. We
looked at our records and 98 of them were for us. We think the military
had the other two. It was a sore point, our own Ultronics modems were
banned and the annual rental on the Post Office's Ericsson things was
triple the retail price of our own, and even then the PO had to modify
every last one of them because they would not immediately shut up when
you dropped RTS, which screwed them for our protocol. You can tell the
pioneers. They are the ones with arrows in their backs!

We also developed a hardware mux back to London from Sydney. We derived
48 teleprinter channels from one 2400 bits/sec leased line. The line
rental was a major fraction of Reuters tech budget.

In 1970 there was very little traffic that was not hardware based. Line
costs were so high that it made economic sense to design protocols and
hardware for the task at hand. Even the original ARPANET had dedicated
hardware devices called TIPs standing in front of each of the PDP-10s
that were slowly giving birth to packet switching. The TIPs eventually
morphed into Honeywell 316's, a very early minicomputer. I remember
attending a seminar by Loius Pouzin on Cyclades and Transpac, which was
one of the progenitors of ARPANET and TCP/IP. Another influential
development was XNS (Xerox Network Services) which made use of the
brand new ethernet hardware locally and PEP and SPP on longer lines.
They were precursors of UDP and TCP respectively. It probably got its
widest use as Novell Netware. Quite a bit later.

There were very few commercial network stacks before the very late
1970s. IBM's SNA and DECnet were the big players, but in many ways they
were quite small. Before V4 of DECnet, (1985-ish) you were limited to
1024 nodes. SNA as originally sold was strictly a star network, with a
requirement for a single node called a SSCP to control the routing.
IBM's internal network was probably the largest in the world, and did
permit peer-peer networking via multiple SSCPs. DECs internal network
was probably the next biggest. Unlike SNA, it permitted a mix of sync
and async lines, but did a much poorer job of polling. I was lucky
enough to get a chance to play with an early prototype of DECnet-11 in
1973, although it was more or less working on PDP-10s a bit before
that.
The major use of DECnet was pretty simple file transfer and getting
your VT100 on one node to pretend it was connected to a remote one.
How many people remember typing SET HOST xxx:: ?
So, apart from Cyclades/Transpac, Telenet (which I didn't mention - it
was a cut-down ARPANET copy) and XNS there was very little commercial
internetworking as we know it today before TCP/IP found its way out of
academia in the early 80's.

As far as local area networks went, in the early 1980s, Ethernet
quickly killed everything else at birth. There were one or two other
ring networks apart from the Cambridge thing, which died a quick and
merciful death. IBM had one, and there was something called ARC. The
group I worked for at DEC had one too, called PCL - a parallel ring
that moved 16 bit words per clock. It was an embarrassment. A lot of
nonsense was spouted about predictability of rings, but the superior
speed of ethernet, and its nobrainer extensibility won the day in short
order. About the only ring network that survives is fibre channel, and
that too is getting killed by 10Gbit ethernet.

There was almost no computer-computer newtworking prior to the late
1960s. Almost every example was connection of a remote teletype to a
computer. At best, remote dumb terminals were multiplexed to a single
host. The multiplexer was always a hardware-only device, until minis
like the PDP-8 became available in the late 1960s. Even they were
wickedly expensive compared to a lump of hardware you could cobble
together from 2N-404s.

--
To de-mung my e-mail address:- fsnospam$elliott$$
PGP Fingerprint: 1A96 3CF7 637F 896B C810 E199 7E5C A9E4 8E59 E248
.

Relevant Pages

  • Re: Wireless modem connection to G4 tower
    ... port to my G4 QS ethernet port via a short ethernet cable (which ... sharing preference panel, click on internet sharing. ... bring up the Network panel showing "Built-in Ethernet", ... cables under rugs in hallways passing into private rooms (where ...
    (comp.sys.mac.apps)
  • Reintroducing... my new PC!
    ... But hardware is only half the computer; ... four days before I got the network straightened out. ... switch futzed out, the internet went down. ... If the proxy computer was ...
    (rec.games.computer.ultima.dragons)
  • Re: Wireless modem connection to G4 tower
    ... port to my G4 QS ethernet port via a short ethernet cable (which ... sharing preference panel, click on internet sharing. ... The only computers that will be able to initiate a connection to the G4 ... Other computers on the existing building Ethernet/WiFi network will not ...
    (comp.sys.mac.apps)
  • Bridging Econet and Ethernet
    ... capable of transferring (via network) files from the internet to old BBC ... What surprises me - even if this is pretty old hardware is that the ethernet ... If not - I guess it's a question of this Ant Internet Suite. ...
    (comp.sys.acorn.networking)
  • Re: HomePlug wired networking
    ... My Pc's upstairs with the modem/router. ... etc. and stringing a load of Ethernet cables blah blah. ... We are now each on the Internet with no pain whatsoever. ... the home network - I am really pleased with it. ...
    (uk.people.silversurfers)