Hacker News Reader

My guess is the telecom vendor your school selected told your school’s IT and purchasing teams that their new phone lines cannot handle modem connections, so they would be obligated to buy a separate IP connectivity service from the same vendor.

Being the early 90s, technical expertise about Internet connectivity was sparse and they most likely entered into a contract with this vendor on the strength of their statement.

They were now surprised to find out the truth: that they could’ve kept using their existing paid-for modems instead of upgrading to a new, expensive, high-speed Internet, access circuit tied to a multi-year contract and all the requisite equipment that came along with that.

The school was likely using ISDN or some sort of digital phone setup and you lucked out. Hotels used to have a special data port on the room handset that you could use because the main line was digital (either ISDN or something else). Even then I could typically only get 28-33k, not the 53k (I never, ever got a pure 56k connection).

Cables are rated for data with a certain expectation of abuse. On a straight run, cat-5 UTP and cat-6 UTP are identical even though the former is rated for 100Mps and the latter for 10Gbps.

But if you bend them around a corner and pull the cable tight, the pairs in the cat-5 will become separated while the pairs in the cat-6 will stay mostly paired (electrons running in opposite directions create fields which negate each other, this is spoiled if the stands separate. Then the cable becomes noisy).

It would be similar with baud rates and phone lines. Probably the school's setup wasn't of the sort that could typically handle data, so the telco had offered them some kind of expensive alternative, but the installers had had a gentle touch, so it actually could handle data.

That, or the telco was just lying and trying to sell something the customer didn't need. Wouldn't be the first time that had happened.

I’m not sure I follow. The service was better than expected?

In my experience you would pay for a minimum service. When we paid for 56k dial-up Bell came in and replaced things until we could achieve 56k. It’s just a noise game. So I imagine it’s possible that the existing lines were decent enough to begin with?

My guess is the IT "department" didn't know the difference between dialup and ISDN.

At least in the schools I did volunteer tech support in the early 00s (in LA, while I was in Middle and High School), the phones in the classrooms were meant to dial into the school’s own switchboard and not to the outside world. Maybe that was possible? My supervising teachers always seemed to walk to the central office to make calls they didn’t want to use their 300 cell phone minutes on (remember that fun?).

The school school had ISDN and then T1 and every computer was networked (even LC IIs) so I don’t think anyone would have bothered trying dialup. In fact, I definitely took advantage of the fast line to download Linux ISOs and other things they were ok with me doing in the off instruction hours.

We even had a small AirPort installation that came care of a donation from Universal Studios or Disney. I had a lot of fun retrofitting an old Rincon 802.11/802.11b card to work with Apple’s Mac OS 8 drivers.

Good memories I am happy to not reproduce given today’s advancements.

This kind of sounds like the line was provided by PBX and they didn't expect modems to work well or at all.

When I was younger, I connected via 56k modem over the PBX at my parents' shop. My connection speeds topped out at 33.6k.

Under the hood, V.90 (the standard for 56k modems) expects that there's only a single digital-to-analog conversion, seen from the ISP's side. The line down from the ISP to the consumer is digital PCM until the last mile. In the opposite direction from the consumer back to the ISP, analog trellis modulation is still used. This is why you still typically only see 33.6k upload speeds on a V.90 modem.

Many PBXes will introduce an additional analog-to-digital conversion in the unit itself, before converting back to analog and putting the signal back on the wire to the street. V.90 can't tolerate that extra conversion, so a connection at 56k speeds fails, and the modem backs down to V.34 / 33.6k, which is perfectly usable on a fully analog line.

This is 20 year old memories so I may have some of the details wrong, but the gist was the school had rolled out an early IP telephony solution and the landlines were just rj-11 patch cables into a monolithic IP telephony box that assigned each line a number and shouldn't have been capable of handling the dial up connection through that

Sounds like an interesting mystery. If a human can talk, a modem can talk. It’s analog after all. Maybe the IP abstraction meant, “there’s quantization happening that makes it technically impossible for a modem to talk at any sensible baud rate. There’s still some circuit switched stuff going on here that we shouldn’t be paying for.”

Now I’m curious to see what baud rate can be achieved over today’s VOIP lines!

I've used analog fax over SIP in the last few years, with ulaw, usually very successfully.

Not sure if it linked at 9600 or what, didn't pay attention.

> If a human can talk, a modem can talk. It’s analog after all.

That’s an oversimplification, especially for 56k where the signal isn’t analog in the sense of the slower speeds and uses PAM (pulse amplitude modulation) that is trained at the beginning of the call (it can be retrained but that is long and drawn out). Central offices had to be upgraded and basic u-law quantization meant 56k was a no go even on most classic POTS lines. There’s a bit more to it than just “noise.”

For VoIP, Amplitude distortion and quantization error (from multiple sources) means you will never get 56k over any voip system. Phase distortion and echo cancellation make other speeds frustrating. Doing better than 9600 is going to be difficult.

My guess is the OP was not dealing with an IP system (sounds less likely for a school in 2000). In any event they noted they could not achieve 56k which is expected that they were not able to establish a PAM link over the PBX, but the channel was good enough for analog FSK modulation. That they could have achieved 28k pretty much rules out any IP system with an ATA of that vintage .

Way back when my dad had an analog POTS connection at his rural Maine house. It worked fine for voice calls but it was flaky to non-existent for a modem connection, to which the telco basically went <shrug>.

If you go back in time, you want to test it with faxing, if it doesn't work for data modem, it probably doesn't work well for faxing, and if it doesn't work for faxing, the telco might actually fix it.

> For VoIP, Amplitude distortion and quantization error (from multiple sources) means you will never get 56k over any voip system

https://frank.petril.li/posts/dialup-adventures-1/

I've done V.90 over VoIP. In a very controlled environment that was tuned for it, but VoIP nonetheless. The main issue is that VoIP timing, even with deep buffers, is prone to more jitter than a PDH network and the phase drift will eventually cause enough errors to force a retrain every few minutes. Either way, "never" is a little too absolute IMHO. :)

Fair enough. How about V.92? Have you had a setup where upstream PCM could be successfully established?

Good question - it's been almost two years since this project so I don't recall whether V.92 came up at all, even on a pure T-1 setup. I still have all the gear, if I set it up and run it again I'll update this thread.

What gear were you using?

https://frank.petril.li/posts/dialup-adventures-1/ explains it all!

Any chance you would be bringing that modem bank to ToorCamp on Orcas Island? The ShadyTel network could always use a better, faster dialup ISP!

I think this is exactly right for reasons that weren't important to the base story. The year after the story they moved to these Cisco VoIP phones that plugged into Ethernet directly, which seems odd considering they had just recently moved onto what was supposedly a Telco provided IP phone service. I suspect someone either misunderstood what they were buying or someone misrepresented what they were selling. They moved to what they thought they were getting in the first place and dropped the telco entirely.

For schools of that time, there was an incredible amount of money flowing around in the dot-com era. At least for a high school in a state capital. Both Microsoft and Cisco sponsored classes and teachers for them at my school. I got semesters A and B of my CCNA as a credit giving elective. My first time using Linux was (ironically) in the Microsoft sponsored class.

Its a big deal if a school computer network that wasn't connected to the internet suddenly you open up a direct route with no firewall or security.

I know it wasn't the same everywhere, but back in 2001 my high school had an IT staff that consisted of a single teacher with no real technical literacy to begin with. There was no real sort of firewall or security, including the computers that were on the internet. It was kind of a different time.

I can hazzard a guess you weren't around back then. Computers weren't something everyone depended on... They were novelties