On 11/07/2021 10:49, MB wrote:
> On 10/07/2021 11:31, Roland Perry wrote:
>> There was probably one in the carry case. I've still got several BNC
>> gubbins in a drawer. In practice very few people used BNC, because it
>> was too topologically fragile. One broken connection and the whole
>> network was likely to go down.
>
>
> It depended a lot on whether you used 50 Ω or 75 Ω BNCs.
>
> Video is 75 Ω but it was standard to use 50 Ω connectors because they
> have a thick centre pin which is less likely to get broken.
>
> (I think I have that the right way around!)
>
>
>
The problem with using 50 Ω connectors in a 75 Ω sockets is the hole in
the socket becomes enlarged and the proper connector no longer fits
properly.

It was a nightmare when colleagues at work started muddling video and
network leads and T pieces and we had expensive video equipment with 75
Ω damaged sockets. Conversely we had network problems with people, in
particular, using 75 Ω.

I remember they had a network problem in a lab I wasn't allowed access
to. The manager and I surmised the problem and decided the cure would
be maximum inconvenience to teach them a lesson so they were instructed
to remove network cards from computers and hand them to us along with
all the cables. We confiscated all the 75 Ω cables and t pieces and
didn't return a few network cards with damaged looking sockets.

They were left to order their own replacements.

On 11/07/2021 11:25, martin.coffee@round-midnight.org.uk wrote:
> The problem with using 50 Ω connectors in a 75 Ω sockets is the hole in
> the socket becomes enlarged and the proper connector no longer fits
> properly.

You very rarely see any 75 Ω connectors.

Similar problem with C connectors, the 75 Ω ones are quite fragile so SC
tended to be used instead.

(Again I think I have got these the right way around! But memory hazy).

In message <scee4a$tqj$1@dont-email.me>, at 10:37:51 on Sun, 11 Jul
2021, MB <MB@nospam.net> remarked:
>On 10/07/2021 10:26, Tweed wrote:
>> It’s amazing how far we’ve come in such a short time. We’ve gone from
>> almost no networking to the planet being covered in it, right down to
>> fairly sophisticated domestic stuff, in less than the length of my working
>> career.
>
>I remember Guy Kewney writing a piece years ago about how originally
>you had to take your work to the IT people (I think they had a
>different name then).

At ICL installations they were "The [computer] Operators". A completely
different set of people fixed the hardware when it went wrong, and yet
another completely separate set of people fixed operating software bugs.

To some extent, the engineers were the most rounded, because they had to
be able to put on an 'Operators' hat, and work the Opper (slang for
Operator's Console), and occasionally write some fragments of ad-hoc
test software.

One of Guy's first jobs was as a programmer for one of the precursors to
ICL. He then went on to start writing about computers in trade
publications such as "Computing" which were originally aimed at
mainframe programmers, and later transitioned to writing for personal
computer publications aimed at the wider public.

>They would run when it suited them so had a lot of power. Then
>computers got more powerful and people could do the jobs themselves,
>taking power away from IT.

The main technology change was from batch processing to timesharing.
Thus users could launch their own jobs from a terminal, and see the
results on screen, rather than it being entirely hands-off. You often
had to go to some shared/centralised facility to get a printout, though.

More powerful, cheaper, and most importantly smaller and more reliable,
computers meant that timesharing mini-computers spread like wildfire. In
most ways a microprocessor-based computer is just an ultra-low-cost
one-person minicomputer (although for far too long people tried largely
in vain to implement timesharing microprocessor installations, to both
multi-serial-terminal users and occasionally multi-memory-mapped-VDU
users).

>Then IT retaliated by having the software and files on the main server
>so they got control again......

Or on a cloud of them, be that on an Intranet or The Internet.

For a while the support nightmare was the way PC users might be allowed
to pick whichever software they wanted to run, without having applied
sufficient critical thought as to its reliability or compatibility (let
alone value for money). That's what made things like Microsoft Office
inevitable - at last the IT department could force everyone to adopt a
"one size WILL fit all" solution, warts and all.
--
Roland Perry

In message <sceeql$2dj$1@dont-email.me>, at 10:49:46 on Sun, 11 Jul
2021, MB <MB@nospam.net> remarked:
>On 10/07/2021 11:31, Roland Perry wrote:

>> There was probably one in the carry case. I've still got several BNC
>> gubbins in a drawer. In practice very few people used BNC, because it
>> was too topologically fragile. One broken connection and the whole
>> network was likely to go down.
>
>It depended a lot on whether you used 50 0 >
>Video is 75 0 >have a thick centre pin which is less likely to get broken.
>
>(I think I have that the right way around!)

I hadn't noticed the difference in the centre pin before , but now you
mention it... the networking ones look as if they have a slightly
thinner pin, and it's gold plated, unlike the a video/audio ones.
--
Roland Perry

On 11/07/2021 11:40, Roland Perry wrote:
> Or on a cloud of them, be that on an Intranet or The Internet.
>
> For a while the support nightmare was the way PC users might be allowed
> to pick whichever software they wanted to run, without having applied
> sufficient critical thought as to its reliability or compatibility (let
> alone value for money). That's what made things like Microsoft Office
> inevitable - at last the IT department could force everyone to adopt a
> "one size WILL fit all" solution, warts and all.

A university friend went to ICL when they recruited a lot of people
straight from university to get their new machine designed and built -
it was way behind schedule.

He was head hunted by someone he knew and went off to a new company in
the USA where he became CTO (I think) and presumably very rich!

"MB" <MB@nospam.net> wrote in message news:sceeql$2dj$1@dont-email.me...
> On 10/07/2021 11:31, Roland Perry wrote:
>> There was probably one in the carry case. I've still got several BNC
>> gubbins in a drawer. In practice very few people used BNC, because it
>> was too topologically fragile. One broken connection and the whole
>> network was likely to go down.
>
>
> It depended a lot on whether you used 50 Ω or 75 Ω BNCs.
>
> Video is 75 Ω but it was standard to use 50 Ω connectors because they have
> a thick centre pin which is less likely to get broken.
>
> (I think I have that the right way around!)

The main problem with BNC T-and-terminator networks was not dodgy
connectors. It was the very topology: you couldn't insert a new T-piece
and/or length of cable without "breaking" the LAN integrity for everyone on
that segment (ie between the terminator at one end and the one at the
other).

In the department where I worked at ICL, we had "the company LAN" which was
cabled with a single T piece to every desk. No-one was allowed to touch that
cable, except to undo the side branch of the T piece from the card in your
PC. That was used for PC access to company services - email, SMB shares to
company servers etc.

Then there was our own team's "private LAN" which went between our desks and
our various servers in the "lab" which we shared with other teams in the
same Network Products Centre. We could do what we liked with that LAN -
insert new T pieces and segments for additional computers etc. There was a
standard protocol which involved going to every person in the team, and if
they weren't at their desk, finding them in the lab, to say "I need to break
the LAN - is it OK with you".

Some network cards and some network protocols were a lot more fussy than
others about unplugging a T piece from a card, while still maintaining the
terminator-to-terminator integrity of the LAN. It should have been perfectly
permissible, but some cards, especially over ICL's OSI-based "OSLAN", threw
a wobbly and maybe even caused a kernel panic on a UNIX server elsewhere on
the LAN. TCP was a lot more tolerant.

Then it was announced that the company would be changing to "structured
cabling" - a cable (or several) from every desk (or every server in the lab)
back to a central hub/switch. Far more UTP cable than we'd needed coaxial
cable in the days of thin Ethernet, but far more reliable because your could
unplug an RJ45 without affecting anyone else on the LAN.

That just left software problems like duplicate IP addresses (in the days
when everyone had a copy of a master HOSTS file, before DHCP), "flooding the
LAN" with a lot of traffic, or the cardinal sin: setting up a duplicate DHCP
server.

In a later job we were all working on developing solutions that ran on
servers. For some reason, the servers that were configured for customers
always ran their own DHCP server (as opposed to using one in a router) and
this always allocated addresses in the 10.0.x.y subnet as opposed to the
normal company standard of 192.168.x.y. Woe betide anyone who accidentally
connected a fully-functional customer-build server onto the company LAN,
because the next PC in the company which requested an IP address might get
allocated a 10.0.x.y address instead of a 192.168.x.y one, and so wouldn't
be able to talk. Maybe the 10/192 distinction was made so that if it *did*
happen, there was never a chance of two computers being given the *same*
address: a useless one is better than a duplicate one. The module that I was
working on involved using Norton Ghost so a PC could request a disc image
from a server if it needed to be restored to a sane state after customers
(schoolchildren) had cocked it up. By default, Ghost client and server
modules were set to communicate flat-out. That would saturate a 10 Mbps LAN,
and would achieve about 90% usage on a 100 Mbps LAN. I only did it once, and
was not very popular ;-) It led me to contact Norton and get them to build
in a new feature: a "throttling" parameter than could be set on the server
module so all Ghost traffic was limited to n Mbps, leaving a plenty of
capacity for all other traffic.

On Sat, 10 Jul 2021 11:31:39 +0100, Roland Perry <roland@perry.co.uk>
wrote:

>In message <scbong$3jh$1@dont-email.me>, at 09:19:44 on Sat, 10 Jul
>2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>Roland Perry <roland@perry.co.uk> wrote:
>>>
>>> ps Laplink cables - that takes me back. Still got one from the 80's,
>>> although I would struggle to find PCs with Centronics printer and/or
>>> 25-pin serial ports to hook it up with. Later I moved to using a
>>> printer-port Ethernet dongle, which was a lifesaver when laptops
>>> didn't have built-in Ethernet, nor any way to add one internally.
>>>
>>> Even had a BNC connctor too!
>
>>>
>>><https://manualzz.com/doc/1009129/technical-bulletin-de-620-ethernet-po
>>>cket-adapter-trouble>
>>
>>Did you carry a T-piece and a teminator just in case?
>
>There was probably one in the carry case. I've still got several BNC
>gubbins in a drawer. In practice very few people used BNC, because it
>was too topologically fragile. One broken connection and the whole
>network was likely to go down.

BNC was actually very common - the IBM 3270 series of display terminal
for their mainframes used them, with coaxial cables from the
controllers.

A bent pin wouldn't take the whole system down, just the controller
and all the (up to) 32 tubes on it.

Roland Perry wrote:

>The main technology change was from batch processing to timesharing.
>Thus users could launch their own jobs from a terminal, and see the
>results on screen, rather than it being entirely hands-off. You often
>had to go to some shared/centralised facility to get a printout, though.

I recall loading some calculations(1) onto a mainframe in
Stafford, via card input in Trafford Park, in the early 70s.
Rather than wait a day for the results to come back, we then
drove to Preston, where we could see them printed out on a
teletype, at roughly the speed of the football results. Don't ask
why this was the case.

The next time, a couple of years later, I had to do a similar
calculation, we used a Compucorp programmable calculator(2).

(1) Start resistor calculations for the Class 313 EMU. I think I
may still have some hand plots of the curves in the loft.

(2) I think this was it:
<http://www.rskey.org/detail.asp?manufacturer=Compucorp&model=326+Scientist>

Chris
--
Chris J Dixon Nottingham UK
chris@cdixon.me.uk @ChrisJDixon1

Plant amazing Acers.

In message <scem4t$ha6$1@dont-email.me>, at 12:54:41 on Sun, 11 Jul
2021, MB <MB@nospam.net> remarked:
>On 11/07/2021 11:40, Roland Perry wrote:
>> Or on a cloud of them, be that on an Intranet or The Internet.
>> For a while the support nightmare was the way PC users might be
>>allowed
>> to pick whichever software they wanted to run, without having applied
>> sufficient critical thought as to its reliability or compatibility (let
>> alone value for money). That's what made things like Microsoft Office
>> inevitable - at last the IT department could force everyone to adopt a
>> "one size WILL fit all" solution, warts and all.
>
>A university friend went to ICL when they recruited a lot of people
>straight from university to get their new machine designed and built -
>it was way behind schedule.

There's a whole bunch of my friends who had been recruited by ICL
Beaumont to be operating system programming interns for what's now
known as a gap year. Then given some kind of bursary to complete their
studies. As far as I know only one went on to go into that industry,
and he was employed at IBM Hursley!

I worked at Letchworth and Bracknell about five years after graduating,
mainly on "New Range" hardware - where the strapline of "Your Future
System" prompted many to say "we'd actually prefer it was here today,
thanks very much".

But getting back to the smaller/cheaper/more_reliable hardware theme,
the very first 2950 (which passed for a minicomputer) out of the factory
was delivered to the engineering training school [you can't deliver to
customers until you have some site engineers ready to fix it when it
goes wrong] was wheeled through the door, plugged in, and worked first
time.

Lots of people present that you could have knocked down with a feather.

>He was head hunted by someone he knew and went off to a new company in
>the USA where he became CTO (I think) and presumably very rich!

That sort of thing happened a lot back then. I don't think the USA-ians
had pulled the immigration ladder up after themselves (Green Cards etc)
quite as much by then. "Brain drain" it was called, not that the UK did
much to try to alleviate it.
--
Roland Perry

In message <scemnl$l6o$1@dont-email.me>, at 13:03:59 on Sun, 11 Jul
2021, NY <me@privacy.invalid> remarked:

>In a later job we were all working on developing solutions that ran on
>servers. For some reason, the servers that were configured for
>customers always ran their own DHCP server (as opposed to using one in
>a router) and this always allocated addresses in the 10.0.x.y subnet as
>opposed to the normal company standard of 192.168.x.y.

Returning to the geolocation theme, I've seen Hollywood cops and robbers
dramas were they've finally tracked down the offender's IP address,
displayed as 192.168.x.y, and set off with blues and twos to somewhere
the other side of town indicated on a map with a flashing pin.
--
Roland Perry

In message <3ksleglur5sv27eve6aou09g9b5p33up0c@4ax.com>, at 14:33:00 on
Sun, 11 Jul 2021, Chris J Dixon <chris@cdixon.me.uk> remarked:
>Roland Perry wrote:
>
>>The main technology change was from batch processing to timesharing.
>>Thus users could launch their own jobs from a terminal, and see the
>>results on screen, rather than it being entirely hands-off. You often
>>had to go to some shared/centralised facility to get a printout, though.
>
>I recall loading some calculations(1) onto a mainframe in
>Stafford, via card input in Trafford Park, in the early 70s.
>Rather than wait a day for the results to come back, we then
>drove to Preston, where we could see them printed out on a
>teletype, at roughly the speed of the football results. Don't ask
>why this was the case.
>
>The next time, a couple of years later, I had to do a similar
>calculation, we used a Compucorp programmable calculator(2).
>
>(1) Start resistor calculations for the Class 313 EMU. I think I
>may still have some hand plots of the curves in the loft.
>
>(2) I think this was it:
><http://www.rskey.org/detail.asp?manufacturer=Compucorp&model=326+Scientist>

$1300, maybe two months salary for a newly qualified engineer at the
time.

Four-function calculators were around £25, the equivalent to about a
week's average wages.
--
Roland Perry

In message <3ksleglur5sv27eve6aou09g9b5p33up0c@4ax.com>, at 14:33:00 on
Sun, 11 Jul 2021, Chris J Dixon <chris@cdixon.me.uk> remarked:
>Roland Perry wrote:
>
>>The main technology change was from batch processing to timesharing.
>>Thus users could launch their own jobs from a terminal, and see the
>>results on screen, rather than it being entirely hands-off. You often
>>had to go to some shared/centralised facility to get a printout, though.
>
>I recall loading some calculations(1) onto a mainframe in
>Stafford, via card input in Trafford Park, in the early 70s.
>Rather than wait a day for the results to come back, we then
>drove to Preston, where we could see them printed out on a
>teletype, at roughly the speed of the football results. Don't ask
>why this was the case.

The speed? That's about as fast as you'd expect it to be.

I loved it when years later the BBC had an onscreen teletype emulator,
which went the same speed, together with accompanying "clatter". When
the results were obviously fully available before it even started the
process.

>The next time, a couple of years later, I had to do a similar
>calculation, we used a Compucorp programmable calculator(2).
>
>(1) Start resistor calculations for the Class 313 EMU. I think I
>may still have some hand plots of the curves in the loft.
>
>(2) I think this was it:
><http://www.rskey.org/detail.asp?manufacturer=Compucorp&model=326+Scientist>
>
>Chris

--
Roland Perry

On 11/07/2021 14:31, Roland Perry wrote:
>> He was head hunted by someone he knew and went off to a new company in
>> the USA where he became CTO (I think) and presumably very rich!
> That sort of thing happened a lot back then. I don't think the USA-ians
> had pulled the immigration ladder up after themselves (Green Cards etc)
> quite as much by then. "Brain drain" it was called, not that the UK did
> much to try to alleviate it.

It was someone he worked with who formed a company. They flew him and
his wife over to look at schools, houses etc.

I rang about a week later for a chat

Asked if he was home (he was often away)

No

When will he be home?

Don't know but hope he is here by the weekend because the removal van
is coming on Monday.

Roland Perry <roland@perry.co.uk> wrote:
> In message <sccner$qdt$1@dont-email.me>, at 18:04:11 on Sat, 10 Jul
> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>> Roland Perry <roland@perry.co.uk> wrote:
>>> In message <scbong$3jh$1@dont-email.me>, at 09:19:44 on Sat, 10 Jul
>>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>>> Roland Perry <roland@perry.co.uk> wrote:
>>>>>
>>>>> ps Laplink cables - that takes me back. Still got one from the 80's,
>>>>> although I would struggle to find PCs with Centronics printer and/or
>>>>> 25-pin serial ports to hook it up with. Later I moved to using a
>>>>> printer-port Ethernet dongle, which was a lifesaver when laptops
>>>>> didn't have built-in Ethernet, nor any way to add one internally.
>>>>>
>>>>> Even had a BNC connctor too!
>>>
>>>>>
>>>>> <https://manualzz.com/doc/1009129/technical-bulletin-de-620-ethernet-po
>>>>> cket-adapter-trouble>
>>>>
>>>> Did you carry a T-piece and a teminator just in case?
>>>
>>> There was probably one in the carry case. I've still got several BNC
>>> gubbins in a drawer. In practice very few people used BNC, because it
>>> was too topologically fragile. One broken connection and the whole
>>> network was likely to go down.
>>
>> When we started installing ethernet across the University at the end of the
>> 1980s we used thin ethernet because that’s what there was - running thick
>> ethernet or long AUI cables through offices was really not an option. And
>> yes, people did do the wrong thing with T-pieces (and Apple blithely
>> invented a transceiver that would happily break a network) and terminators
>> (we found one member of staff who used to take the terminator of the cable
>> in his office and fiddle with it as a displacement activity; I can’t
>> remember whether we gave him one of his own as a desk toy).
>>
>> “An Ethernet is not only a distributed packet switch, it is also a
>> distributed single point of failure.” Boggs, Mogul and Kent, 1988, IIRC.
>
> <https://www.informationweek.com/computer-glitch-strands-thousands-at-
> lax-what-went-wrong/d/d-id/1058148?piddl_msgorder=thrd&>
>
> Nine hours to find what I think other reports confirmed as a broken
> Ethernet card in just one PC, crunching the whole internal network.

Interesting. I’m sure we had similar events, and what you do is start
chopping off bits of the network until it works again, and then put them
back until it breaks. In 2007 I can’t remember to what extent we’d have
been able to do that remotely rather than send a techician in to pull
cables out.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

In message <scf8ak$h8e$1@dont-email.me>, at 17:04:20 on Sun, 11 Jul
2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>Roland Perry <roland@perry.co.uk> wrote:
>> In message <sccner$qdt$1@dont-email.me>, at 18:04:11 on Sat, 10 Jul
>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>> Roland Perry <roland@perry.co.uk> wrote:
>>>> In message <scbong$3jh$1@dont-email.me>, at 09:19:44 on Sat, 10 Jul
>>>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>>>> Roland Perry <roland@perry.co.uk> wrote:
>>>>>>
>>>>>> ps Laplink cables - that takes me back. Still got one from the 80's,
>>>>>> although I would struggle to find PCs with Centronics printer and/or
>>>>>> 25-pin serial ports to hook it up with. Later I moved to using a
>>>>>> printer-port Ethernet dongle, which was a lifesaver when laptops
>>>>>> didn't have built-in Ethernet, nor any way to add one internally.
>>>>>>
>>>>>> Even had a BNC connctor too!
>>>>
>>>>>>
>>>>>> <https://manualzz.com/doc/1009129/technical-bulletin-de-620-ethernet-po
>>>>>> cket-adapter-trouble>
>>>>>
>>>>> Did you carry a T-piece and a teminator just in case?
>>>>
>>>> There was probably one in the carry case. I've still got several BNC
>>>> gubbins in a drawer. In practice very few people used BNC, because it
>>>> was too topologically fragile. One broken connection and the whole
>>>> network was likely to go down.
>>>
>>> When we started installing ethernet across the University at the end of the
>>> 1980s we used thin ethernet because that’s what there was - running thick
>>> ethernet or long AUI cables through offices was really not an option. And
>>> yes, people did do the wrong thing with T-pieces (and Apple blithely
>>> invented a transceiver that would happily break a network) and terminators
>>> (we found one member of staff who used to take the terminator of the cable
>>> in his office and fiddle with it as a displacement activity; I can’t
>>> remember whether we gave him one of his own as a desk toy).
>>>
>>> “An Ethernet is not only a distributed packet switch, it is also a
>>> distributed single point of failure.” Boggs, Mogul and Kent, 1988, IIRC.
>>
>> <https://www.informationweek.com/computer-glitch-strands-thousands-at-
>> lax-what-went-wrong/d/d-id/1058148?piddl_msgorder=thrd&>
>>
>> Nine hours to find what I think other reports confirmed as a broken
>> Ethernet card in just one PC, crunching the whole internal network.
>
>Interesting. I’m sure we had similar events, and what you do is start
>chopping off bits of the network until it works again, and then put them
>back until it breaks. In 2007 I can’t remember to what extent we’d have
>been able to do that remotely rather than send a techician in to pull
>cables out.

I don't think t-pieces are remotely [on any sense of the word] able to
be individually managed like that.
--
Roland Perry

Tweed <usenet.tweed@gmail.com> wrote:
> MB <MB@nospam.net> wrote:
>> On 10/07/2021 10:26, Tweed wrote:
>>> It’s amazing how far we’ve come in such a short time. We’ve gone from
>>> almost no networking to the planet being covered in it, right down to
>>> fairly sophisticated domestic stuff, in less than the length of my working
>>> career.
>>
>> I remember Guy Kewney writing a piece years ago about how originally you
>> had to take your work to the IT people (I think they had a different
>> name then). They would run when it suited them so had a lot of power.
>> Then computers got more powerful and people could do the jobs
>> themselves, taking power away from IT.
>>
>> Then IT retaliated by having the software and files on the main server
>> so they got control again......
>>
>>
>
> That’s so true. These days they are using GDPR and security as reasons for
> reasserting their grip. To some extent it is justified, but it’s certainly
> a gift to the central IT departments.

It’s also a matter of corporate governance, knowledge and information
management, and business resilience. Society has changed as well as the
technology, so while it’s now easier for people to take information around
with them (and mishandle it, too, obvs) it’s also true that businesses now
have much more of their processes invested in IT systems and any loss or
mishandling can have much graver consequences.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

Roland Perry <roland@perry.co.uk> wrote:
> In message <scf8ak$h8e$1@dont-email.me>, at 17:04:20 on Sun, 11 Jul
> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>> Roland Perry <roland@perry.co.uk> wrote:
>>> In message <sccner$qdt$1@dont-email.me>, at 18:04:11 on Sat, 10 Jul
>>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>>> Roland Perry <roland@perry.co.uk> wrote:
>>>>> In message <scbong$3jh$1@dont-email.me>, at 09:19:44 on Sat, 10 Jul
>>>>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>>>>> Roland Perry <roland@perry.co.uk> wrote:
>>>>>>>
>>>>>>> ps Laplink cables - that takes me back. Still got one from the 80's,
>>>>>>> although I would struggle to find PCs with Centronics printer and/or
>>>>>>> 25-pin serial ports to hook it up with. Later I moved to using a
>>>>>>> printer-port Ethernet dongle, which was a lifesaver when laptops
>>>>>>> didn't have built-in Ethernet, nor any way to add one internally.
>>>>>>>
>>>>>>> Even had a BNC connctor too!
>>>>>
>>>>>>>
>>>>>>> <https://manualzz.com/doc/1009129/technical-bulletin-de-620-ethernet-po
>>>>>>> cket-adapter-trouble>
>>>>>>
>>>>>> Did you carry a T-piece and a teminator just in case?
>>>>>
>>>>> There was probably one in the carry case. I've still got several BNC
>>>>> gubbins in a drawer. In practice very few people used BNC, because it
>>>>> was too topologically fragile. One broken connection and the whole
>>>>> network was likely to go down.
>>>>
>>>> When we started installing ethernet across the University at the end of the
>>>> 1980s we used thin ethernet because that’s what there was - running thick
>>>> ethernet or long AUI cables through offices was really not an option. And
>>>> yes, people did do the wrong thing with T-pieces (and Apple blithely
>>>> invented a transceiver that would happily break a network) and terminators
>>>> (we found one member of staff who used to take the terminator of the cable
>>>> in his office and fiddle with it as a displacement activity; I can’t
>>>> remember whether we gave him one of his own as a desk toy).
>>>>
>>>> “An Ethernet is not only a distributed packet switch, it is also a
>>>> distributed single point of failure.” Boggs, Mogul and Kent, 1988, IIRC.
>>>
>>> <https://www.informationweek.com/computer-glitch-strands-thousands-at-
>>> lax-what-went-wrong/d/d-id/1058148?piddl_msgorder=thrd&>
>>>
>>> Nine hours to find what I think other reports confirmed as a broken
>>> Ethernet card in just one PC, crunching the whole internal network.
>>
>> Interesting. I’m sure we had similar events, and what you do is start
>> chopping off bits of the network until it works again, and then put them
>> back until it breaks. In 2007 I can’t remember to what extent we’d have
>> been able to do that remotely rather than send a techician in to pull
>> cables out.
>
> I don't think t-pieces are remotely [on any sense of the word] able to
> be individually managed like that.

That article was published in 2007. Twisted pair ethernet was standardised
in 1990. If LAX was susceptible to a 9 hour outage in 2007 because it was
still using T-pieces it really only had itself to blame. Whether or not a
T-piece (or more realistically a switch or repeater port connecting to a
coax segment, the kind that needs T-pieces) was remotely manageable is
hardly relevant.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

<martin.coffee@round-midnight.org.uk> wrote:
> On 11/07/2021 10:49, MB wrote:
>> On 10/07/2021 11:31, Roland Perry wrote:
>>> There was probably one in the carry case. I've still got several BNC
>>> gubbins in a drawer. In practice very few people used BNC, because it
>>> was too topologically fragile. One broken connection and the whole
>>> network was likely to go down.
>>
>>
>> It depended a lot on whether you used 50 Ω or 75 Ω BNCs.
>>
>> Video is 75 Ω but it was standard to use 50 Ω connectors because they
>> have a thick centre pin which is less likely to get broken.
>>
>> (I think I have that the right way around!)

Thin Ethernet is indeed 50 Ω.

> The problem with using 50 Ω connectors in a 75 Ω sockets is the hole in
> the socket becomes enlarged and the proper connector no longer fits
> properly.

Which would be a problem when you tried to use it for video but it would
still work for ethernet.

> It was a nightmare when colleagues at work started muddling video and
> network leads and T pieces and we had expensive video equipment with 75
> Ω damaged sockets. Conversely we had network problems with people, in
> particular, using 75 Ω.
>
> I remember they had a network problem in a lab I wasn't allowed access
> to. The manager and I surmised the problem and decided the cure would
> be maximum inconvenience to teach them a lesson so they were instructed
> to remove network cards from computers and hand them to us along with
> all the cables. We confiscated all the 75 Ω cables and t pieces and
> didn't return a few network cards with damaged looking sockets.

Was that just pique? Network cards with 50 Ω connectors shouldn’t have a
problem with 75 Ω T-pieces or cables, right?

> They were left to order their own replacements.

:-)

We once had a DEC engineer come in to look at our VAX and declare that it
was currently unsupportable. We asked why. He said that the ethernet
cable was the wrong colour, it was black and not grey. We protested that
it was certified RG-58 equivalent. Yes, he said, but DEC didn’t support
equivalent cable, only genuine RG-58 which is specified with a grey sheath.

So we swapped the piece of cable that was visibly attached to the VAX.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

MB <MB@nospam.net> wrote:
> On 11/07/2021 11:25, martin.coffee@round-midnight.org.uk wrote:
>> The problem with using 50 Ω connectors in a 75 Ω sockets is the hole in
>> the socket becomes enlarged and the proper connector no longer fits
>> properly.
>
>
> You very rarely see any 75 Ω connectors.

I used to work in an ultrasound research group, long before I got into
networking. All our coax cable was 75 Ω because we were working with
video.

> Similar problem with C connectors, the 75 Ω ones are quite fragile so SC
> tended to be used instead.
>
> (Again I think I have got these the right way around! But memory hazy).

We used BNC exclusively (there’s a nice picture of the difference between
50 Ω and 75 Ω at <https://en.wikipedia.org/wiki/BNC_connector#Types>) but a
hospital research group isn’t a particularly onerous environment.

SC connectors are for optical fibre. When you say SC are you thinking of
SMC, which are like a smaller, screw version of a BNC? C are like a
larger, more robust-looking BNC. I don’t know which would have worked best
in your environment.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

Roland Perry <roland@perry.co.uk> wrote:
> In message <sceeql$2dj$1@dont-email.me>, at 10:49:46 on Sun, 11 Jul
> 2021, MB <MB@nospam.net> remarked:
>> On 10/07/2021 11:31, Roland Perry wrote:
>
>>> There was probably one in the carry case. I've still got several BNC
>>> gubbins in a drawer. In practice very few people used BNC, because it
>>> was too topologically fragile. One broken connection and the whole
>>> network was likely to go down.
>>
>> It depended a lot on whether you used 50 0 >
>> Video is 75 0 >have a thick centre pin which is less likely to get broken.
>>
>> (I think I have that the right way around!)
>
> I hadn't noticed the difference in the centre pin before , but now you
> mention it... the networking ones look as if they have a slightly
> thinner pin, and it's gold plated, unlike the a video/audio ones.

I’ve just mentioned in another sub-thread, there’s a good image at
<https://en.wikipedia.org/wiki/BNC_connector#Types>.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

NY <me@privacy.invalid> wrote:
> "MB" <MB@nospam.net> wrote in message news:sceeql$2dj$1@dont-email.me...
>> On 10/07/2021 11:31, Roland Perry wrote:
>>> There was probably one in the carry case. I've still got several BNC
>>> gubbins in a drawer. In practice very few people used BNC, because it
>>> was too topologically fragile. One broken connection and the whole
>>> network was likely to go down.
>>
>>
>> It depended a lot on whether you used 50 Ω or 75 Ω BNCs.
>>
>> Video is 75 Ω but it was standard to use 50 Ω connectors because they have
>> a thick centre pin which is less likely to get broken.
>>
>> (I think I have that the right way around!)
>
> The main problem with BNC T-and-terminator networks was not dodgy
> connectors. It was the very topology: you couldn't insert a new T-piece
> and/or length of cable without "breaking" the LAN integrity for everyone on
> that segment (ie between the terminator at one end and the one at the
> other).

Unless you were very clever, like Apple, who invented an in-line
transceiver with built in terminators and microswitches. That meant you
didn’t have to move the terminator when you extended a cable, just plug an
extra length on. But that meant you had to break it to insert the
transceiver rather than just plug it into the T-piece, unless you were
clever enough to just plug one of the connectors on the transceiver into
the T-piece, which introduced an extra terminator in the segment and
usually broke it for everyone. The cure was to add yet another T-piece on
the other socket to switch out the terminator.

> In the department where I worked at ICL, we had "the company LAN" which was
> cabled with a single T piece to every desk. No-one was allowed to touch that
> cable, except to undo the side branch of the T piece from the card in your
> PC. That was used for PC access to company services - email, SMB shares to
> company servers etc.
>
> Then there was our own team's "private LAN" which went between our desks and
> our various servers in the "lab" which we shared with other teams in the
> same Network Products Centre. We could do what we liked with that LAN -
> insert new T pieces and segments for additional computers etc. There was a
> standard protocol which involved going to every person in the team, and if
> they weren't at their desk, finding them in the lab, to say "I need to break
> the LAN - is it OK with you".
>
> Some network cards and some network protocols were a lot more fussy than
> others about unplugging a T piece from a card, while still maintaining the
> terminator-to-terminator integrity of the LAN. It should have been perfectly
> permissible, but some cards, especially over ICL's OSI-based "OSLAN", threw
> a wobbly and maybe even caused a kernel panic on a UNIX server elsewhere on
> the LAN. TCP was a lot more tolerant.

When I left the University about 18 months ago we still had corporate
systems where the front end talked to a back end database using a protocol
that timed out in about 3 seconds, at which point the whole system had to
be restarted. Getting network maintenance slots was a nightmare and even
quite usual disturbances that most applications wouldn’t notice at all
caused these important corporate applications to crash completely. Utterly
bizarre.

> Then it was announced that the company would be changing to "structured
> cabling" - a cable (or several) from every desk (or every server in the lab)
> back to a central hub/switch. Far more UTP cable than we'd needed coaxial
> cable in the days of thin Ethernet, but far more reliable because your could
> unplug an RJ45 without affecting anyone else on the LAN.
>
> That just left software problems like duplicate IP addresses (in the days
> when everyone had a copy of a master HOSTS file, before DHCP), "flooding the
> LAN" with a lot of traffic, or the cardinal sin: setting up a duplicate DHCP
> server.
>
> In a later job we were all working on developing solutions that ran on
> servers. For some reason, the servers that were configured for customers
> always ran their own DHCP server (as opposed to using one in a router) and
> this always allocated addresses in the 10.0.x.y subnet as opposed to the
> normal company standard of 192.168.x.y. Woe betide anyone who accidentally
> connected a fully-functional customer-build server onto the company LAN,
> because the next PC in the company which requested an IP address might get
> allocated a 10.0.x.y address instead of a 192.168.x.y one, and so wouldn't
> be able to talk. Maybe the 10/192 distinction was made so that if it *did*
> happen, there was never a chance of two computers being given the *same*
> address: a useless one is better than a duplicate one. …

Apple, I’m looking at you again. >-( AirPort, Time Capsule, various
things contained a DHCP server, and when a lecturer[1] brought one in
because he liked the way his wireless network worked at home, they could
bring a whole building down with a kind of creeping paralysis. The
solution usually ended up with us configuring a central monitoring port on
the affected VLAN and looking for rogue DHCP traffic.

[1] Corporate technical standards? Ha! A colleague referred to the
University as a franchise for 9000 sole traders.

> … The module that I was
> working on involved using Norton Ghost so a PC could request a disc image
> from a server if it needed to be restored to a sane state after customers
> (schoolchildren) had cocked it up. By default, Ghost client and server
> modules were set to communicate flat-out. That would saturate a 10 Mbps LAN,
> and would achieve about 90% usage on a 100 Mbps LAN. I only did it once, and
> was not very popular ;-) It led me to contact Norton and get them to build
> in a new feature: a "throttling" parameter than could be set on the server
> module so all Ghost traffic was limited to n Mbps, leaving a plenty of
> capacity for all other traffic.

Ah, yes, Ghost. IP multicast without IGMP. That meant the network
equipment couldn’t know which hosts needed the traffic and which didn’t, so
it had to send it everywhere. When you had a mixed speed network, either
different speed equipment or hosts that negotiated different speeds, and a
server which was capable of over 10 Mbps any 10 Mbps hosts were just
flattened, whether they were taking part in the Ghost session or not.
Terrible design.

Sam

--
The entity formerly known as Sam.Wilson@ed.ac.uk
Spit the dummy to reply

In message <scfobm$rke$1@dont-email.me>, at 21:37:58 on Sun, 11 Jul
2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>Roland Perry <roland@perry.co.uk> wrote:
>> In message <scf8ak$h8e$1@dont-email.me>, at 17:04:20 on Sun, 11 Jul
>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>> Roland Perry <roland@perry.co.uk> wrote:
>>>> In message <sccner$qdt$1@dont-email.me>, at 18:04:11 on Sat, 10 Jul
>>>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>>>> Roland Perry <roland@perry.co.uk> wrote:
>>>>>> In message <scbong$3jh$1@dont-email.me>, at 09:19:44 on Sat, 10 Jul
>>>>>> 2021, Sam Wilson <ukr@dummy.wislons.fastmail.co.uk> remarked:
>>>>>>> Roland Perry <roland@perry.co.uk> wrote:
>>>>>>>>
>>>>>>>> ps Laplink cables - that takes me back. Still got one from the 80's,
>>>>>>>> although I would struggle to find PCs with Centronics printer and/or
>>>>>>>> 25-pin serial ports to hook it up with. Later I moved to using a
>>>>>>>> printer-port Ethernet dongle, which was a lifesaver when laptops
>>>>>>>> didn't have built-in Ethernet, nor any way to add one internally.
>>>>>>>>
>>>>>>>> Even had a BNC connctor too!
>>>>>>
>>>>>>>>
>>>>>>>> <https://manualzz.com/doc/1009129/technical-bulletin-de-620-ethernet-po
>>>>>>>> cket-adapter-trouble>
>>>>>>>
>>>>>>> Did you carry a T-piece and a teminator just in case?
>>>>>>
>>>>>> There was probably one in the carry case. I've still got several BNC
>>>>>> gubbins in a drawer. In practice very few people used BNC, because it
>>>>>> was too topologically fragile. One broken connection and the whole
>>>>>> network was likely to go down.
>>>>>
>>>>> When we started installing ethernet across the University at the
>>>>>end of the
>>>>> 1980s we used thin ethernet because that’s what there was -
>>>>>running thick
>>>>> ethernet or long AUI cables through offices was really not an option. And
>>>>> yes, people did do the wrong thing with T-pieces (and Apple blithely
>>>>> invented a transceiver that would happily break a network) and terminators
>>>>> (we found one member of staff who used to take the terminator of the cable
>>>>> in his office and fiddle with it as a displacement activity; I can’t
>>>>> remember whether we gave him one of his own as a desk toy).
>>>>>
>>>>> “An Ethernet is not only a distributed packet switch, it is also a
>>>>> distributed single point of failure.” Boggs, Mogul and Kent,
>>>>>1988, IIRC.
>>>>
>>>> <https://www.informationweek.com/computer-glitch-strands-thousands-at-
>>>> lax-what-went-wrong/d/d-id/1058148?piddl_msgorder=thrd&>
>>>>
>>>> Nine hours to find what I think other reports confirmed as a broken
>>>> Ethernet card in just one PC, crunching the whole internal network.
>>>
>>> Interesting. I’m sure we had similar events, and what you do is start
>>> chopping off bits of the network until it works again, and then put them
>>> back until it breaks. In 2007 I can’t remember to what extent we’d have
>>> been able to do that remotely rather than send a techician in to pull
>>> cables out.
>>
>> I don't think t-pieces are remotely [on any sense of the word] able to
>> be individually managed like that.
>
>That article was published in 2007. Twisted pair ethernet was standardised
>in 1990. If LAX was susceptible to a 9 hour outage in 2007 because it was
>still using T-pieces it really only had itself to blame.

Big organisations (especially public sector and/or in the USA) are
notorious for using out-of-date infrastructure, a combination of lack of
funds and make-do-and-mend culture that keeps things rumbling along, but
ever more difficult to repair or replace.

It's highly likely the cabling was in ducts where it's impossible to
tell where the other end of any particular one going in, comes out. And
often half the cables aren't connected to *anything* any more because
rushed repair jobs in the past just run a new cable rather than try to
fix the old one.

Although the article also blames what amounts to a disorganised IT team
(who were on-call, not on-site), and I expect like you they also had
problems finding routine maintenance slots in what's effectively a 24x7
operation.

In any event, they first had to localise where the fault was, initially
suspecting some ISP routers:

With remote diagnostics showing no problems after three hours
and mounting delays at the airport, Sprint sent technicians to
the site. Around 8:00 p.m., it became clear that the problem was
not with the network connection but with CBP's local area
network.

After that, they must have started unplugging bits of the internal
network to isolate the fault, because terminals 2, 4, 5 & 7 were up and
running fairly soon after, it was another (where presumably the trigger
fault was) that took longer.
--
Roland Perry

On 11/07/2021 23:11, Sam Wilson wrote:
> SC connectors are for optical fibre. When you say SC are you thinking of
> SMC, which are like a smaller, screw version of a BNC? C are like a
> larger, more robust-looking BNC. I don’t know which would have worked best
> in your environment.

No, 'Screwed C' (SC) RF connectors. Screwed version of "N" connector
just as the TNC is the screwed version of BNC.

On 11/07/2021 23:11, Sam Wilson wrote:
> I’ve just mentioned in another sub-thread, there’s a good image at
> <https://en.wikipedia.org/wiki/BNC_connector#Types>.

ISA (Inter-Series Adaptor I think) adaptor kits have 50 Ω and 75 Ω
connectors, I think the insulation / spacer in the 75 Ω ones is red.
Will check one later.

"Roland Perry" <roland@perry.co.uk> wrote in message
news:ScRdGB6DLv6gFAJ4@perry.uk...
> There's a whole bunch of my friends who had been recruited by ICL Beaumont

ICL Beaumont (at Old Windsor) was a fantastic building. It had been a Jesuit
college until (I think) the 1960s, and consisted of lots of different wings
all built at different times on different levels, so there were "secret
passages" and "hidden staircases". I first went there when I was on my
teens, in the long hot summer of 1976, on a course run for schools by ICL's
CES (Computer Education in Schools) department.

What we learned was fairly mind-blowing by 1970s schoolboy standards (though
not by 2021 standards!) - being able to issue instructions to computers in a
language called CESIL (Computer Education in Schools Instructive Language)
or in BASIC, mostly by writing on squared coding sheets which were punched
onto tape, taken to a mainframe (maybe in Bracknell), run, and the printout
returned to us the following day.

And after classes, we explored the building, going down all the secret
passages and hidden staircases (I remember there was a long spiral one). We
found the big war memorial in the grounds, dedicated to the staff and boys
of the Jesuit college who had died in the two world wars.

One day we were taken by mini bus to another building about half an hour's
drive away, which had a first-floor gallery with wired-glass windows that
looked down on a huge computer hall with big cabinets which it was explained
were processor units, tape drives, disk drives, printers and teletype
consoles.

Fast forward about 15 years. I got a job at ICL BRA01 (all ICL sites were
named with three letters to denote the town [Bracknell, in this case] and
two digits to distinguish different buildings in the town). I'd been there
several years when I had to go down to a different department and saw the
"goldfish bowl" windows. So *that* was where we'd been taken all those years
ago.

I had to give a few training courses at Beaumont, and on one occasion they
had some builders in, doing some work on a ceiling. And through the missing
tiles in the false ceiling I could look up into what had been the chapel
until it had extra floors and rooms inserted, and the sunlight was streaming
through a magnificent stained-glass window.