If you ever want more phone lines than that, you can pick up an old Cisco VG-224 from Ebay for less than half the price of that line simulator, and you get 24 lines. There is a configuration that will let you use it as a standalone unit where all the lines can call each other with custom phone numbers (here's some notes [1]).
The main catch is that they have a 50-pin Centronics style connector on them which you will have to break out somehow to your RJ11s. Also, they are big (1U rack) and have fans.
I've got a few of these and have been meaning to set them up with a bunch of modems and a bunch of computers, but haven't gotten to it yet. Modems do seem to work in the limited testing I've done. They do (as expected) work great with telephones, including pulse dialing.
Do the VG224's manage V.90/V.92 speeds? Jeff's setup falls into the usual trap other YouTubers do. Dialing a modem through an ATA or other device directly into another modem caps the speeds at V.34 (33600 baud)
From what I understand of telecom of that era. You want to effectively keep as much of the signal digital as possible. With (ideally) the only "analog" part remaining being the link between the customer modem and the PBX.
The VG224 being (effectively) 24 ATA's in a trenchcoat seems to meet that requirement. Though once installed you'd need to connect it to a PBX such as Asterisk. Then, as I understand it, "trunk" that as a T1 line into something that can digitally handle modem calls such as a Cisco AS5300.
> The main catch is that they have a 50-pin Centronics style connector on them which you will have to break out somehow to your RJ11s.
This is just another lost art (traditional phones are either dead or are instead IP) that I once learned a fair bit about:
The name varies regionally (I've heard them called Centronics, cinch, and CHAMP; though around here, we call them Amphenols). The Easy Method is the same regardless of name: It centers around a split (aka "50 pair") 66 punch block[1] that is mounted to a wall, or to a wall-like object.
Buy a pre-terminated 25-pair cable with the right connector on at least one end, and punch that down in order[2] on one side of the 66 block. That connects the system to the punch block. Importantly, those wires never get touched again.
Phones (or more precisely, wires for jacks for phones) connect to the other side of the 66 block. Those wires also never get touched again.
The two things (phones, systems) are connected/disconnected with bridge clips that combine the two halves of the block (which only allows 1:1 ordering, but that's often just fine).
Alternatively, a "we fancy!" variation uses single-pair cross-connect wire so that arbitrary phones can quickly be connected to arbitrary system ports -- maybe on completely separate blocks.
After that, plug in the Amphenol. Plug in the phones. Have fun talking to yourself.
(Or, at least: That's an easy way for small stuff. Bigger stuff (hundreds or thousands of pairs) eventually really wants better organization, but punch blocks are still normally the order of the day there, too.)
I know we ripped out similar, a plastic panel/wall mount version, at an old job when we went VoIP. The Centronics came out of the PBX. Probably find on eBay or surplus. Maybe talk to IT and they have one hanging around in a junk pile.
Sure. One end of the 25-pair cable can be cut down on a patch panel. There doesn't need to be a 66 block.
Or a person can buy a patch panel pre-assembled with an Amphenol connector, and just plug in a 25-pair cable between that and their gear. Here's one from Leviton: https://leviton.com/products/49012-j24
And those kinds of things are great for rapid deployment, but we're not trying to do that -- are we? We're just goofing around with old telco stuff for the lulz.
The most tinkering-friendly option is 66 block (wherein: the structure is whatever we want it to be today), I think. Punch down tools are cheap (including those that have interchangeable blades for also doing tidy work of 110 keystone jacks). It only takes a few minutes to land all 25 pairs, and those pairs can do anything a person wants them to do.
Meanwhile, 66 blocks themselves: They're dandy things. I've used them in the field to build simple circuits: The resistor or the capacitor doesn't care that it is hanging off of a punch down block instead of soldered to a PCB. It works, and it's wall-mountable for permanence.
But yeah, there's certainly options. I've got good familiarity with 66 blocks and their ability to quickly (and rather universally) hack stuff together, so that's what I tend towards using.
It's always important to know the specific and most-correct nomenclature, but it's equally important to use terms that others easily understand. :)
Like, the 4-pin Molex connectors in common use for things like PC optical drives (where those still exist). You probably know exactly what I'm referring to, and can picture one in your head whenever someone mentions a "Molex connector" in the context of a PC.
But those aren't from Molex -- they're instead from Amphenol's Mate-n-Lok line. Molex sells compatible-enough versions (as do countless other manufactures), but Molex had nothing to do with the design or introduction of that part.
Meanwhile: The square 4/6/8/20/24-pin jobbies that connect to things like video cards and motherboards, and that seemingly everyone calls ATX connectors? Those actually are Molex parts (from their Mini-Fit Jr line), but people would be confused AF if you called them Molex connectors.
So if we're doing it right in casual discourse, we use the terms that people understand -- even if they're not necessarily maximally-correct.
I've never witnessed the term "micro ribbon connector" used in face-to-face dealings involving telco stuff.
---
Up next: The male 1/4" audio connector. Is it a headphone plug? A guitar plug? TR? TRS? A phone plug? A jack plug? "Like an aux cable but bigger?" Yes, any of those terms work. The best term depends on the audience.
The connector originates in the forgotten world of plug-and-socket telephone switchboards, so of all of those options "phone plug" probably gets the history most-correct (and that's also what Radio Shack labeled them as). But that specific terminology is pretty archaic at this point.
It's so archaic that if I asked a sane and rational person for a cable with "phone plug" on it these days, they might go looking for something with 6P4C, 8P8C, USB C, or (maybe!) a TRS -- depending on their age and/or upbringing.
The other catch is actually subscribing to those phone lines (real POTS lines, not VOIP lines). Where I live, the only game in town is Bell Canada. The cost for a single home phone line is $58/month!
They do have enterprise accounts where I presume you'd be able to subscribe to 24 phone lines, but that would not be cheap! Whether they'd even allow you to bring 24 phone lines into a residential house is another question. They might not even have trunk capacity to offer you that many lines at your residence, so then you'd need to lease office space so they could bring in a T1 line.
Ahh I see. I guess I was mistaken about the title, which suggests building a real ISP for friends & family, rather than just a simulated ISP within a local network.
Semi-related note:I learned this morning that there are companies that offer in-home (non-cellular) phone service via wifi/the internet. Think: classic cordless phone, however the base station connects via wifi. Some of those companies offer free unlimited service as well...and the one I was looking at has a companion app for iOS/Android that allows you to make calls from via home number from anywhere as long as you have an internet connection.
I'm assuming these numbers are flagged as VOIP (which limits their use as a lot of apps/sites/companies hate VOIP numbers), however the discover is interesting. I thought home phones were dead and buried.
I had a talk about building your own dial-up ISP in FOSDEM 2026, not the best but it remarks some points and configuration tips for especially running on VoIP. Thanks for keeping dial-up alive!
Unfortunately they're not plug-and-play _yet_ but people have significantly streamlined the process of making dialup bridges with Raspis and USB modems [1] to restore online multiplayer on the Sega Dreamcast (which came with a 56k modem)
I've wondered for a while if it would be possible to do that with a USB FXS adapter[1], and something like a Pi, but you'd basically need a soft modem of some kind to make it work and the only module for Asterisk isn't up to the challenge (last I checked five years ago).
However, to be frank, it'd make more sense to do PPP over null modem with a straight serial connection.
I'd be curious if anyone has actually used a setup like this as a fallback for when their fiber goes down. The latency would be brutal, but a few kbps beats no bps
Jeff mentioned in his video that just loading the front page of CNN would take something like an hour and a half (20+ MB).
33.6Kbps is not practical for much on the modern Internet in 2026. As mentioned in a sibling comment, Starlink (even in standby mode) would be much better. lite.cnn.com would load in about 10 seconds which is pretty good, but there's not much else like it left anymore.
What's amazing is how great the Internet in the 1990s managed to be despite these limitations. Just like with RAM and disk space, developers back then had to be very mindful of bandwidth - today's devs (and agents) have the luxury of paying much less attention to that.
20+ mb is also the weight of rendering the HTML inside each client, instead of at the server. It is the weight of continuous disdain for users, and of 30 years of not giving a fuck about adding yet another abstraction layer and making it someone else's problem.
You can somewhat "fix" this by using your slow link to connect to a VPS somewhere that then connects to the Internet, either via links or a similar text-mode browser, or other bandwidth-saving gateway.
In the article and my pi-isp project, I use MacProxy Classic to strip heavy stuff from web pages through a local proxy service running on the Pi. This helps a lot, but if a page has 20 MB of resources, it can only do so much (without completely disabling JS and images).
There were proxies in the old day that would recompress images and do other things, but if the 20 MB is compressed javascript, there's not much you can do but hope it caches well.
That's a lot of stuff that I don't want to deal with or maintain. It's simply beyond the tolerance of my gumption, so it's not going to happen. :)
What can happen, instead: I can dream.
In this dream, the process of loading a web page identifies the viewing platform well-enough and the server delivers content that is shaped for it, so it can be downloaded quickly and displayed simply by the end-user device. It's not one-size-fits-all at all, or even one-size-fits-most: It's a pile of of simplistic HTML and maybe some minimal javascript and CSS that is meant for whatever the user is using right now.
In this way, the same layout jiggering, varnicating, and transfabulation is done as it is done today, but the work of doing so principally happens on the server instead of the client.
Also in this dream, I can hear people saying "But that's a can of worms!", and they're right. It's a damned mess -- but it's a mess either way. This just moves the mess from the client to the server.
I can also hear shouts of "But there will be hundreds or even thousands of layout paths!" And all I can think is: If there's a thousand unique device types hitting a given dynamic page, and that scales poorly with the server side doing the work, then that's a problem for the systems guys to direct instead of the web guys.
Which is fine: The web guys hacking away however they want is how we got into this mess of 20 megabyte Javascript downloads just-to-view-a-web-page to begin with. They've quite broadly proven that they're shit at this kind of work, and in my ideal world they'd be relieved of that duty.
(And yeah, to be sure: After I wake from this dream I'm still going to go outside and yell at the clouds, just as I do every day.)
This dream actually existing for a hot minute back in the day! There were a number of sites that would give you much different experiences based on your capabilities - and this continued into the mobile era - until Jobs screwed it up by shipping a desktop-capable web browser on a phone.
I pulled out a WRT54G the other day and set it to a 56K limit. A netbook running ubuntu(ish) took several days to complete an apt-get update. But when I connected my laptop with 50 browser tabs, the tabs themselves consumed 100% of the bandwidth with background traffic (despite my ad blocker) and no other network services worked at all.
Unless you've tuned your system for it, dial-up modem speeds are functionally equivalent to "no connection at all".
Timeouts are killer too... many modern apps assume slow or no response for more than 10 or 20 seconds is "Internet is down" and will stop trying.
It's brutal, even with modern Internet, when people develop apps with the assumption of the Internet connection always having decent latency and bandwidth.
You have to get a dish on a different plan first but then you can switch to standby mode, it's $5/mo. It might not be available if you rent the hardware though, you might have to buy the dish outright.
I feel like it would have been more fun to build your own line simulator with a 9 volt battery and some old phone line and just skip the dialtone altogether for a little magic... the $120 black-box telco simulator takes a lot of the fun out.
But then again, based on Pi pricing today, the $120 telco simulator goes nicely with a $300 Pi 5.
Line simulators are fun. The ports can call each other and there's no need to set any custom init strings on the modem like you need with the 9 volt trick. For some old devices that's a necessity.
You can also use an old VoIP ATA from Linksys/Cisco as a cheap line simulator. Like a fully analog TLS the ports can call each other. They can be a PITA to configure right but they're cheap and work well enough.
I've used all three methods, the TLS is the easiest. An ATA can be useful if you've got more than one and your dial-in server is in a different room from the client you're playing with. An ATA can also be set to "call" another device. So your office ATA can call the basement ATA (with your Linux server) as an example.
Just issue ATX0D on one modem and ATA on the other. Bingo bango: One modem thinks it's initiating a call (without concern about the lack of dialtone), and the other thinks it is answering a call.
But yeah, simulators are fun. A person I used to call "boss" used one with a dry pair of copper that ran all the way across town for his internet connection for a couple of years.
(His shop had space that was used by a local dialup ISP for a POP. In exchange for the space and the electricity, his shop got free internet over a T1 in the days before we had DSL or DOCSIS. It was a great trade for both parties and I felt giddy downloading things at work at >1.5 Mbps instead of ~0.0336 Mbps.
In an effort to save some money on a dedicated phone line and ISP bill at home, he ordered a cheap circuit from the phone company that they billed as an "alarm circuit" -- it was about $20 per month.
One modem was his at house behaving mostly-normally along with the simulator, and another was at the shop connected to our all-singing, all-dancing 486DX4 Linux box. It was stable and cheap and somewhat ridiculous.)
If you ever want more phone lines than that, you can pick up an old Cisco VG-224 from Ebay for less than half the price of that line simulator, and you get 24 lines. There is a configuration that will let you use it as a standalone unit where all the lines can call each other with custom phone numbers (here's some notes [1]).
The main catch is that they have a 50-pin Centronics style connector on them which you will have to break out somehow to your RJ11s. Also, they are big (1U rack) and have fans.
I've got a few of these and have been meaning to set them up with a bunch of modems and a bunch of computers, but haven't gotten to it yet. Modems do seem to work in the limited testing I've done. They do (as expected) work great with telephones, including pulse dialing.
[1] https://alnwlsn.com/z/pots/cisco-vg224.html
Do the VG224's manage V.90/V.92 speeds? Jeff's setup falls into the usual trap other YouTubers do. Dialing a modem through an ATA or other device directly into another modem caps the speeds at V.34 (33600 baud)
From what I understand of telecom of that era. You want to effectively keep as much of the signal digital as possible. With (ideally) the only "analog" part remaining being the link between the customer modem and the PBX.
The VG224 being (effectively) 24 ATA's in a trenchcoat seems to meet that requirement. Though once installed you'd need to connect it to a PBX such as Asterisk. Then, as I understand it, "trunk" that as a T1 line into something that can digitally handle modem calls such as a Cisco AS5300.
> The main catch is that they have a 50-pin Centronics style connector on them which you will have to break out somehow to your RJ11s.
This is just another lost art (traditional phones are either dead or are instead IP) that I once learned a fair bit about:
The name varies regionally (I've heard them called Centronics, cinch, and CHAMP; though around here, we call them Amphenols). The Easy Method is the same regardless of name: It centers around a split (aka "50 pair") 66 punch block[1] that is mounted to a wall, or to a wall-like object.
Buy a pre-terminated 25-pair cable with the right connector on at least one end, and punch that down in order[2] on one side of the 66 block. That connects the system to the punch block. Importantly, those wires never get touched again.
Phones (or more precisely, wires for jacks for phones) connect to the other side of the 66 block. Those wires also never get touched again.
The two things (phones, systems) are connected/disconnected with bridge clips that combine the two halves of the block (which only allows 1:1 ordering, but that's often just fine).
Alternatively, a "we fancy!" variation uses single-pair cross-connect wire so that arbitrary phones can quickly be connected to arbitrary system ports -- maybe on completely separate blocks.
After that, plug in the Amphenol. Plug in the phones. Have fun talking to yourself.
(Or, at least: That's an easy way for small stuff. Bigger stuff (hundreds or thousands of pairs) eventually really wants better organization, but punch blocks are still normally the order of the day there, too.)
[1]: https://en.wikipedia.org/wiki/66_block
[2]: https://news.ycombinator.com/item?id=41428998
Instead of 66 blocks and punch down tools how about something like this: https://www.showmecables.com/icc-telco-male-patch-panel-rj11...
I know we ripped out similar, a plastic panel/wall mount version, at an old job when we went VoIP. The Centronics came out of the PBX. Probably find on eBay or surplus. Maybe talk to IT and they have one hanging around in a junk pile.
> The name varies regionally
Micro Ribbon connector https://en.wikipedia.org/wiki/Micro_ribbon_connector
Sure. One end of the 25-pair cable can be cut down on a patch panel. There doesn't need to be a 66 block.
Or a person can buy a patch panel pre-assembled with an Amphenol connector, and just plug in a 25-pair cable between that and their gear. Here's one from Leviton: https://leviton.com/products/49012-j24
Or, there's unholy creations like this octopus breakout cable, which can be used to connect existing patch panels/structured cable plants into Amphenol world: https://www.sandman.com/products/cbl3c-6-foot-qwik-25x2-cabl...
And those kinds of things are great for rapid deployment, but we're not trying to do that -- are we? We're just goofing around with old telco stuff for the lulz.
The most tinkering-friendly option is 66 block (wherein: the structure is whatever we want it to be today), I think. Punch down tools are cheap (including those that have interchangeable blades for also doing tidy work of 110 keystone jacks). It only takes a few minutes to land all 25 pairs, and those pairs can do anything a person wants them to do.
Meanwhile, 66 blocks themselves: They're dandy things. I've used them in the field to build simple circuits: The resistor or the capacitor doesn't care that it is hanging off of a punch down block instead of soldered to a PCB. It works, and it's wall-mountable for permanence.
But yeah, there's certainly options. I've got good familiarity with 66 blocks and their ability to quickly (and rather universally) hack stuff together, so that's what I tend towards using.
> Micro Ribbon connector https://en.wikipedia.org/wiki/Micro_ribbon_connector
It's always important to know the specific and most-correct nomenclature, but it's equally important to use terms that others easily understand. :)
Like, the 4-pin Molex connectors in common use for things like PC optical drives (where those still exist). You probably know exactly what I'm referring to, and can picture one in your head whenever someone mentions a "Molex connector" in the context of a PC.
But those aren't from Molex -- they're instead from Amphenol's Mate-n-Lok line. Molex sells compatible-enough versions (as do countless other manufactures), but Molex had nothing to do with the design or introduction of that part.
Meanwhile: The square 4/6/8/20/24-pin jobbies that connect to things like video cards and motherboards, and that seemingly everyone calls ATX connectors? Those actually are Molex parts (from their Mini-Fit Jr line), but people would be confused AF if you called them Molex connectors.
So if we're doing it right in casual discourse, we use the terms that people understand -- even if they're not necessarily maximally-correct.
I've never witnessed the term "micro ribbon connector" used in face-to-face dealings involving telco stuff.
---
Up next: The male 1/4" audio connector. Is it a headphone plug? A guitar plug? TR? TRS? A phone plug? A jack plug? "Like an aux cable but bigger?" Yes, any of those terms work. The best term depends on the audience.
The connector originates in the forgotten world of plug-and-socket telephone switchboards, so of all of those options "phone plug" probably gets the history most-correct (and that's also what Radio Shack labeled them as). But that specific terminology is pretty archaic at this point.
It's so archaic that if I asked a sane and rational person for a cable with "phone plug" on it these days, they might go looking for something with 6P4C, 8P8C, USB C, or (maybe!) a TRS -- depending on their age and/or upbringing.
The other catch is actually subscribing to those phone lines (real POTS lines, not VOIP lines). Where I live, the only game in town is Bell Canada. The cost for a single home phone line is $58/month!
They do have enterprise accounts where I presume you'd be able to subscribe to 24 phone lines, but that would not be cheap! Whether they'd even allow you to bring 24 phone lines into a residential house is another question. They might not even have trunk capacity to offer you that many lines at your residence, so then you'd need to lease office space so they could bring in a T1 line.
I don’t think anyone is suggesting getting actual phone lines, but rather more lines in a local network setting.
Ahh I see. I guess I was mistaken about the title, which suggests building a real ISP for friends & family, rather than just a simulated ISP within a local network.
Semi-related note:I learned this morning that there are companies that offer in-home (non-cellular) phone service via wifi/the internet. Think: classic cordless phone, however the base station connects via wifi. Some of those companies offer free unlimited service as well...and the one I was looking at has a companion app for iOS/Android that allows you to make calls from via home number from anywhere as long as you have an internet connection.
I'm assuming these numbers are flagged as VOIP (which limits their use as a lot of apps/sites/companies hate VOIP numbers), however the discover is interesting. I thought home phones were dead and buried.
Most ISPs for the last 10 years or so offer this. Many modems come with a POTS RJ11 jack on them.
Yes, it's VOIP. You might get a "real" number with appropriate area code that might route through the real POTS network. Depends on locale and ISP
So basically VOIP?
I had a talk about building your own dial-up ISP in FOSDEM 2026, not the best but it remarks some points and configuration tips for especially running on VoIP. Thanks for keeping dial-up alive!
https://fosdem.org/2026/schedule/event/UJKT3L-dial-up-howto/
What I would love to have is a few compact dialup-to-wifi bridges so I could wifi-ize 30 year old hardware.
Would be neat to read email with an old POP client, or chat over the original AIM software (perhaps patched to use a server on the LAN)
Unfortunately they're not plug-and-play _yet_ but people have significantly streamlined the process of making dialup bridges with Raspis and USB modems [1] to restore online multiplayer on the Sega Dreamcast (which came with a 56k modem)
[1] https://dreamcastlive.net/dreampi-tutorial/
I've wondered for a while if it would be possible to do that with a USB FXS adapter[1], and something like a Pi, but you'd basically need a soft modem of some kind to make it work and the only module for Asterisk isn't up to the challenge (last I checked five years ago).
However, to be frank, it'd make more sense to do PPP over null modem with a straight serial connection.
[1]: https://www.amfeltec.com/usb-fxs-adapter/
I'd be curious if anyone has actually used a setup like this as a fallback for when their fiber goes down. The latency would be brutal, but a few kbps beats no bps
Jeff mentioned in his video that just loading the front page of CNN would take something like an hour and a half (20+ MB).
33.6Kbps is not practical for much on the modern Internet in 2026. As mentioned in a sibling comment, Starlink (even in standby mode) would be much better. lite.cnn.com would load in about 10 seconds which is pretty good, but there's not much else like it left anymore.
What's amazing is how great the Internet in the 1990s managed to be despite these limitations. Just like with RAM and disk space, developers back then had to be very mindful of bandwidth - today's devs (and agents) have the luxury of paying much less attention to that.
It wasn’t all that great.
Simple weather radar image on Compuserve took about a minute to load. Billing was by the minute too!
What’s more amazing is with 10000x the communication speed and about 1000x the CPU, 1/20th the latency, things now are only about 30x faster…
Today, that damn weather page should have been fully loaded and rendered before one even realized they pressed the enter key.
There are services which will remove all but the text when browsing and make it greatly lighter.
20+ mb is the weight of all the javascript javascripting, ultimately to arrange and display an html page.
20+ mb is also the weight of rendering the HTML inside each client, instead of at the server. It is the weight of continuous disdain for users, and of 30 years of not giving a fuck about adding yet another abstraction layer and making it someone else's problem.
You can somewhat "fix" this by using your slow link to connect to a VPS somewhere that then connects to the Internet, either via links or a similar text-mode browser, or other bandwidth-saving gateway.
In the article and my pi-isp project, I use MacProxy Classic to strip heavy stuff from web pages through a local proxy service running on the Pi. This helps a lot, but if a page has 20 MB of resources, it can only do so much (without completely disabling JS and images).
There were proxies in the old day that would recompress images and do other things, but if the 20 MB is compressed javascript, there's not much you can do but hope it caches well.
That's a lot of stuff that I don't want to deal with or maintain. It's simply beyond the tolerance of my gumption, so it's not going to happen. :)
What can happen, instead: I can dream.
In this dream, the process of loading a web page identifies the viewing platform well-enough and the server delivers content that is shaped for it, so it can be downloaded quickly and displayed simply by the end-user device. It's not one-size-fits-all at all, or even one-size-fits-most: It's a pile of of simplistic HTML and maybe some minimal javascript and CSS that is meant for whatever the user is using right now.
In this way, the same layout jiggering, varnicating, and transfabulation is done as it is done today, but the work of doing so principally happens on the server instead of the client.
Also in this dream, I can hear people saying "But that's a can of worms!", and they're right. It's a damned mess -- but it's a mess either way. This just moves the mess from the client to the server.
I can also hear shouts of "But there will be hundreds or even thousands of layout paths!" And all I can think is: If there's a thousand unique device types hitting a given dynamic page, and that scales poorly with the server side doing the work, then that's a problem for the systems guys to direct instead of the web guys.
Which is fine: The web guys hacking away however they want is how we got into this mess of 20 megabyte Javascript downloads just-to-view-a-web-page to begin with. They've quite broadly proven that they're shit at this kind of work, and in my ideal world they'd be relieved of that duty.
(And yeah, to be sure: After I wake from this dream I'm still going to go outside and yell at the clouds, just as I do every day.)
This dream actually existing for a hot minute back in the day! There were a number of sites that would give you much different experiences based on your capabilities - and this continued into the mobile era - until Jobs screwed it up by shipping a desktop-capable web browser on a phone.
A couple of news sites have a low-bandwidth version: https://text.npr.org
I’ve been thinking about using LLMs as a sort of proxy, for both dial-up and bad phone networks.
Theoretically, such a service could allow you to "browse" even via SMS.
I mean, it’s a bit ridiculous to have to resort to that, but the situation itself is ridiculous.
I pulled out a WRT54G the other day and set it to a 56K limit. A netbook running ubuntu(ish) took several days to complete an apt-get update. But when I connected my laptop with 50 browser tabs, the tabs themselves consumed 100% of the bandwidth with background traffic (despite my ad blocker) and no other network services worked at all.
Unless you've tuned your system for it, dial-up modem speeds are functionally equivalent to "no connection at all".
Timeouts are killer too... many modern apps assume slow or no response for more than 10 or 20 seconds is "Internet is down" and will stop trying.
It's brutal, even with modern Internet, when people develop apps with the assumption of the Internet connection always having decent latency and bandwidth.
It would be funny as a project but there's better low speed backup options like a starlink dish on standby mode (500kbps)
Is that publicly available? And how much would that cost?
You have to get a dish on a different plan first but then you can switch to standby mode, it's $5/mo. It might not be available if you rent the hardware though, you might have to buy the dish outright.
The way this is set up is using your normal fiber internet as the backbone, so if the fiber goes down, the dialup does, too.
I feel like it would have been more fun to build your own line simulator with a 9 volt battery and some old phone line and just skip the dialtone altogether for a little magic... the $120 black-box telco simulator takes a lot of the fun out.
But then again, based on Pi pricing today, the $120 telco simulator goes nicely with a $300 Pi 5.
Line simulators are fun. The ports can call each other and there's no need to set any custom init strings on the modem like you need with the 9 volt trick. For some old devices that's a necessity.
You can also use an old VoIP ATA from Linksys/Cisco as a cheap line simulator. Like a fully analog TLS the ports can call each other. They can be a PITA to configure right but they're cheap and work well enough.
I've used all three methods, the TLS is the easiest. An ATA can be useful if you've got more than one and your dial-in server is in a different room from the client you're playing with. An ATA can also be set to "call" another device. So your office ATA can call the basement ATA (with your Linux server) as an example.
It's not particularly custom:
Just issue ATX0D on one modem and ATA on the other. Bingo bango: One modem thinks it's initiating a call (without concern about the lack of dialtone), and the other thinks it is answering a call.
But yeah, simulators are fun. A person I used to call "boss" used one with a dry pair of copper that ran all the way across town for his internet connection for a couple of years.
(His shop had space that was used by a local dialup ISP for a POP. In exchange for the space and the electricity, his shop got free internet over a T1 in the days before we had DSL or DOCSIS. It was a great trade for both parties and I felt giddy downloading things at work at >1.5 Mbps instead of ~0.0336 Mbps.
In an effort to save some money on a dedicated phone line and ISP bill at home, he ordered a cheap circuit from the phone company that they billed as an "alarm circuit" -- it was about $20 per month.
One modem was his at house behaving mostly-normally along with the simulator, and another was at the shop connected to our all-singing, all-dancing 486DX4 Linux box. It was stable and cheap and somewhat ridiculous.)
Don't need to do that. I use a USR Total Control chassis as a white noise generator.
Nice, but slip and plip is more tun
[dead]