What's that smell ... LANGuard

[rrinker]

 You have to admit though, at least in the thinnet days we had cool t-connector and terminator sculptures on our desks...

[JimS]

It looks like on the standard RJ-45 socket.
The upper right 2 pins have a PCB via attached to them.
Or something copper looking.

As for the eye pictures.
No way to tell what the setup was when the picture was taken.
Could have been anything.
Marketing...gotta love it.

[CJay]

You have to admit though, at least in the thinnet days we had cool t-connector and terminator sculptures on our desks...

'had'?

I've still got a box full of 10B2 stuff, it's handy for quick lashups for HF radios.

[bd139]

I've seen 10base2 still in service in the last year. What's worrying is that it was connected to an AUI box hanging off the back of something that should have been chucked in a skip 20 years ago made by DEC. Insurance sector scares the fuck out of me.

[ebastler]

Am I the only one bothered by the poor grammar on that page?

If you are bothered by (or even aware of) grammar, you are probably not in the target audience. 

[CJay]

I've seen 10base2 still in service in the last year. What's worrying is that it was connected to an AUI box hanging off the back of something that should have been chucked in a skip 20 years ago made by DEC. Insurance sector scares the fuck out of me.

Insurance, Banking, Doorstep Loan companies, all still using DEC stuff and DEC terminals hanging off DECServers, loads of warehouses and factories too, the DEC is short for DECRepit...

There's good reasons why banking IT lags, it costs a *lot* to upgrade and their hands are tied by some arcane rules concerning their IT but it's still scary to work in the data centres of major banks that rely on >20 year old equipment for core business.

[rrinker]

 Since we have a lot of banks as clients - it's almost funny how so many of them have very up to date IT - until you get to the core banking app. Then it's old, hosted stuff running on antique mainframes. Just when you thought you saw the last 3270 terminal emulator - then you see what runs on the teller workstations.
 One of my clients has a division that buys and sells loans. All the funds transfers involved are wire transactions. Until a couple of years ago when I wrote an application for them, they were doing all these wire transfers manually - can be hundreds per day. Crazy. The file format for an automated wire transfer is quite archaic, but is not difficult to create and it is very well documented. Even worse were the incoming ones - someone had to read through the transfer report and then manually update the information in their primary application. Not only was it slow and time consuming, the potential for error was rather high. Now it's all automated, both ways, and they handle many more transactions per day.

 Hmm, I did find some T connectors in the back room, wonder if I have enough to make a desktop pet. I don't do any RF stuff so I never thought to keep that stuff around once everywhere I worked switched to UTP cabling. I think there's a spool of coax in the back too.  Never forget the one time a client called us out for a problem with a workstation connecting to the network. Go to the server/wiring room, and coming out of one of their UTP hubs was this run of flat phone cable. Two of us just looked at each other, and then said to the client "I assume this is the cable going to the workstation with a problem?" Not only was it the wrong type of cable, even if it WAS proper Cat 3 cable, it was far longer than the distance limit with UTP at the time, because it went up some 20 feet to the ceiling, and came down at the workstation end, in addition to being about as far away as you could get in this huge warehouse/manufacturing plant. And freely strung across whatever else was up in the rafters - everything from 120V to 48V power lines, fluorescent lighting fixtures, etc.

[CJay]

Since we have a lot of banks as clients - it's almost funny how so many of them have very up to date IT - until you get to the core banking app. Then it's old, hosted stuff running on antique mainframes. Just when you thought you saw the last 3270 terminal emulator - then you see what runs on the teller workstations.

Until recently we had a contract that involved BRUSYS tapes and 20MB disk drives, PDP11s running the core day to day business of a household name organisation in the UK who's annual turnover is in the billions.

Some of the nuclear stuff runs on computers that are very nearly as old as I am and I'm 50 this year.

[dmills]

Some of the nuclear stuff runs on computers that are very nearly as old as I am and I'm 50 this year.

I knew a guy who made good money off that market, apparently he had a stock of original carbon comp resistors for the bus termination cards, and as they were original spec they could be used without having to touch the safety case paperwork (After all there are no well known issues with 40 year old carbon comp resistors, right?)....

Putting modern metal film in instead would have solved the reliability problem but might have forced redoing several tons of safety paperwork!

Regards, Dan.

[bd139]

I love how these threads descend into lamenting the state of reality

As for carbon comp resistors I was talking to someone I know locally the other week who doesn’t do the internet and he’s got a lot of old components floating around. We’re talking tens of thousands of carbon comp resistors in neat little drawers. All in his damp uninsulated 1950s cheap build with 1970s gas central heating  .  You can see the mould. He’s probably sitting on a fortune to be honest but he’s mostly interested in keeping it out of the hands of “guitar peons” and using it for “real ham radio” which he doesn’t do past being racist on SSB...

[CJay]

Chuck a laptop with a copy of WSJTx at him and see him explode

[bd139]

Hahaha. I’m just going to buy him a ladder 

[Howardlong]

If you’ve not suffered 10base5 or Type 1 token ring cabling, you’ve not lived ;-)

[bd139]

I have once. When I was younger and stupider I tried to pull some CAT5 through with a section of 10base5 (RG8 but worse). It didn’t move. 

[CJay]

Oh I've done Token Ring too, spent many hours pulling that cable, resetting MAUs and making a hash of fitting those godawful plugs.

Have also done Twinax.

[rrinker]

 Yup, way back when, I worked in a token ring shop. In the one building, they decided to make the server/wiring room a show place, with a big glass wall on one side. Trying to neatly route and attach type 1 TR cables to a rack full of MAUs was NOT fun. I think the building used to be owned by IBM, and it was set up that way as a demo room. Which is fine - but we used it for production networking. Fun fun, NOT.
 During that time I was also sent to a token ring course at an IBM office in New York City. Union rules in the city were such that for some of the exercises, the instructor was making statements like "imagine if this cable was removed from the MAU" since no one but a union electrician was allowed to change ANY wiring.

[Howardlong]

That was the main problem with Type 1, keeping the wiring closets and patch panels neat. Token ring was flaky anyway, one bad node and your whole LAN segment died. DOS and Windows 3.1 just didn’t behave well. Finding a bad machine causing beaconing was a crap shoot. Try to repatch something, your whole LAN segment often died, which in itself was why patch panels became so messy, no one dared move anything for fear of taking down a whole floor!

Around 1992 or so, IBM persuaded us to move from MAUs to CAUs and LAMs, which had such crappy firmware we had an even worse network. Of course, no body ever got fired for buying IBM.

This was around the time IBM lost its way in my experience.

We also nad a problem with their PS/2 server RAID configs and drives. First, you’d lose multiple drives simultaneously on an array, especially after a power down. The IBM tech arrived, hammer in hand, tapped the offending drives with the right force, unsticking the heads. The drives were replaced, server boots fine.

Second RAID cock up was to put the config on battery back RAM on the controller card with no means of config backup... which then became a single point of failure.

We switched to Compaq.

[bd139]

I liked Compaq. You can still get the Conpaq hardware experience with HPE now. Just avoid shitty Broadcom NICs

[Bassman59]

I liked Compaq. You can still get the Conpaq hardware experience with HPE now. Just avoid shitty Broadcom NICs

Yeah, but Compaq bought DEC, so watch out!

[CJay]

I liked Compaq. You can still get the Conpaq hardware experience with HPE now. Just avoid shitty Broadcom NICs

Yeah, but Compaq bought DEC, so watch out!

Pfft, DEC and Compaq were and are still both great kit, it's testament to their quality that there are still many Alpha servers and older Compaq Proliant machines running some very heavy duty applications in safety critical systems.

HPE gear is similarly excellent, used to love working on their SANs and servers (I think my accreditations are still valid too)

[bd139]

DEC kit was always good.

HPE is generally quite good. The only turd I've seen from HPE in the last decade is one of the switches they sell for the C7000 blade chassis is buggy as fuck. Inevitably we went and bought a whole stack of them so that was an expensive mistake. Also the NICs on the LOM cards in the DL580 g8 are buggy thus occasionally we have to send in the resident PFY to get felt up by the DC security guys just to replug the damn thing. We're going DAS enterprise SSD and piles of DL380 g10 and redundant Juniper EX on our refresh this year as we killed off the last monolithic monster database servers. Can't wait to inspect the cabling 

[rrinker]

 Yeah, we were an IBM shop (6x 3090 processor complexes, though 3 were the Amdahl clone versions) and of course the token ring - but after a few too many server issues with the PS/2 servers, we switched to the original Compaq Proliants. And yeah, those IBM acronyms - what is this, a network or a barnyard? MAUs, CAUs, LAMs... those were the days. Until they decided the entire IT department was unnecessary, at whch point they went through at least 5 different service providers before realizing that no, you will not get non-employees to some in on a Sunday afternoon and stay all night fixing a server so the business unit can function come Monday morning. It was during one of those times when I got a handle on just how much tape was actually in an 8mm cartridge - while waiting for a server load, I took a bad one and tied the loose end to a doorknob of an office int he center stack of the building, and then ran around unspooling it to see how many times I could circle the floor. It was a lot. 
 And I had it done, AND cleaned up before the server required further input.

[IanMacdonald]

There should be NO earth on any pin of an Ethernet RJ45. If there is, it is potentially unsafe. There are normally four line isolating transformers (one for each pair) between the socket and the signal handling chip. 

The reason it's like this is because Ethernet often has to run between buildings, and there may be earth potential differences of a few volts, but with the capability to drive several hundred amps (Basically the neutral current for the entire building) through any link between them. If the earths of the computers are linked the cable may overheat, or tracks be burned off the PCB.  Used to be a major issue with the earlier coax networks.

Also PoE should be between pairs, not across the conductors of a pair. Provided the conductors have no short to earth, PoE should have no effect on a device which doesn't use it. I suspect the problem here is because somebody thought it would be clever to earth the conductors, or maybe to put in MOVs which can't take the PoE voltage.

[wraper]

There should be NO earth on any pin of an Ethernet RJ45. If there is, it is potentially unsafe. There are normally four line isolating transformers (one for each pair) between the socket and the signal handling chip. 

The reason it's like this is because Ethernet often has to run between buildings, and there may be earth potential differences of a few volts, but with the capability to drive several hundred amps (Basically the neutral current for the entire building) through any link between them. If the earths of the computers are linked the cable may overheat, or tracks be burned off the PCB.  Used to be a major issue with the earlier coax networks.

Also PoE should be between pairs, not across the conductors of a pair. Provided the conductors have no short to earth, PoE should have no effect on a device which doesn't use it. I suspect the problem here is because somebody thought it would be clever to earth the conductors, or maybe to put in MOVs which can't take the PoE voltage.

Then go check actual schematic how transformers can be connected. Voltage between differential pairs most likely will be an issue, thus proper PoE supply should have load detection.
BTW you forgot about shielded cables which are a must for outdoor use.

[lty1993]

If I am right, the 1000BASE-T uses PAM5. The eye diagram they showed are NRZ signal!!!