Hacker News Reader

Comment by antirez | Hacker News Reader

A bit less than 30 years ago my friend and I found an IBM AT in the trash, perfect conditions, with even the floppy disks nearby. We connected it with an RS232 cable to a Linux box, and made an IRC workstation using QMODEM, for the girlfriend of my friend. We were connecting via dialups, and the bill every month was high. With Linux NAT, we were already sharing between my and his apartment, using a coaxial cable, and with the IBM AT now we could share the Internet connection to three computers.

View on HN

Seeing "internet connection" mentioned in connection with QMODEM is weird. Are you sure you got your story right?

Modems were the way to connect to the internet providers so I guess he used QMODEM to dial up.

That's not at all what QMODEM does.

That's precisely what QMODEM does? What do you think it does?

QMODEM didn't connect to the Internet. It had no IP stack.

It could connect you to a machine that had Internet access. Some ISPs offered that as a service (you'd get some kind of BBS-like interface or - if you were lucky - a UNIX shell), but that's not the same thing.

QMODEM was essentially just a terminal emulator that used a serial port and understood how to control a modem.

Exactly.

After all this discussion here my conclusion is that "connecting to the internet" is an ambiguous term.

It can mean "have IP connectivity, i.e., IP packets routed to and from the internet" in which case the described PC was not "connected to the internet with QMODEM". It didn't do anything IP.

It can mean "have a terminal that can interact with information from/to the internet" in which case the PC was indeed "connected to the internet with QMODEM".

To me, the second meaning is quite the stretch, but apparently to others it's fine.

Yes, I think it's a very ambigious term.

My problem with your definition is that it doesn't take into account the reality of connectivity at the time (at least in my experience of the early 90s) - not a whole lot of machines had IP stacks that were connected via ethernet/isdn/t1/etc and online all the time. Certainly you'd have to be pretty special to actually own one or have one at home. Connecting over some kind of tty or dialup was extremely common.

So using your definition, a sizeable percentage (possibly even a majority?) of people who were online and doing things on the internet during the QModem era were doing it through computers that were not "connected to the internet". Which seems obviously silly.

Indeed, I was there, I know. As a starving college student, using QModem for part of it.

> that's not the same thing.

I think your definition of "connect to the internet" make sense today, but would be ridiculously narrow when applied to the QModem era given the computing landscape at the time. Where do you draw the line? Using a tty style terminal connected via serial to a unix box connected via ethernet? How about SLIP/PPP?

I guess my problem with your definition is that you end up saying that a very large percentage of people who were online at the time were using the internet through computers that were not "connected to the internet".

Until the mid-90s the internet was predominantly text anyway, so it's not like you were missing out on a whole lot if you were "only" using a terminal.

My definition of a computer being on the Internet is the computer has an IP stack that can route directly to other networks. In this scenario, the Linux box at the other end of the serial cable was on the Internet. The machine running QModem was not.

However, the user running QModem was on the Internet.

qmodem probably makes a perfectly good terminal program. So presumably the 3rd PC gets its internet connection (so to speak) by being a terminal for one of the other 2.

I'm reading it as they used the IBM AT as a serial terminal, running QMODEM as the terminal emulator, for a Linux host that was connected to the Internet via modem.

Saying "share the internet connection" is quite a stretch though if you just had a terminal connection to some other host which then connected to the internet. I'd associate "sharing a connection" with some (perhaps NATed) IP routing. And they mention NAT, thus my question.

Maybe your association would be different / the terminology would make more sense if you were online in the early 90s?

BUT, it was definitely possible to do what you're describing with some combination of a dialup shell account, a terminal program like qmodem and something like https://en.wikipedia.org/wiki/Slirp

I think running SLiRP on the Linux box and a SLIP client on the IBM AT was probably a stretch, but it's certainly possible. At that point it probably would have made more sense to grab an NE2000 NIC and throw the IBM AT onto the coax network.

Oh I wasn't trying to reverse-engineer his network from that comment, just saying that this was a thing that was possible and that people did at the time.

I agree it's highly unlikely that the AT was running slirp. Wikipedia says an AT was a 286, so it wouldn't have been linux. Not even sure what the options would have been. Minix? Xenix?

Xenix definitely ran on a 286. I can't say re: Minix-- I never have used it (though I probably should just to have the experience). I believe there was a Crynwr SLIP packet driver.

QModem ran on DOS, so the AT wasn't running UNIX. It's almost certainly being used as a terminal.

I can confirm that they did run Minix OK, although I remember the network support was iffy at best. We never got it to work at any rate. XENIX would have been hard to get your hands on. I think QNX would run on an AT as well, although my memory might be playing tricks on me there.

Most network software for DOS was LAN-oriented, like Novell or NetBIOS. Just drive mapping and printer redirection. I'm not aware of a TCP/IP connectivity suite being available for DOS in that era, and I'm not sure how it would have worked given that DOS provided no networking libraries to hook into.

I was online in the 90s and I did have a 286 PC. Thus my questions. (That PC was running DOS and didn't have much of a chance to have IP connectivity.)

Based on their mention of "coax" I bet they have a Linux box w/ a modem doing dial-up PPP to an ISP, and a 10Base-2 NIC that they used to attach another PC. The Linux box was doing IP masquerading (NAT) to share the PPP connection w/ the machine(s) on the 10Base-2 LAN.

Having the IBM AT a a serial terminal would let somebody run CLI-based software on the Linux box (like Lynx, an IRC client, FTP, etc). You'd just be using a shell account on the Linux box.

I did stuff like this in the early 90s at home and later at a company I worked at (sharing a single dial-up connection over 10Base-2 with 5-ish Windows 95 PCs).

My first "real internet connection" (not prodigy or compuserve or local BBS) was connecting to the county library's gopher server via modem (Likely with QModem), navigating out to other gopher servers and then being able to telnet from there.

The clue is in the rs232 mentioned in my comment. Linux is a Unix, even if nobody does it today you can configure an external terminal via serial port and use the system as if you were a user.

The two Linux box were one with the modem and the other via Nat (Ethernet but with old coaxial cables). The AT was just a terminal.

Having a serial terminal was just good fun, too, back in the days when having multiple monitors / computers wasn't common. I ran Slackware on a 486 w/ X Windows so I could use Netscape Navigator and I had an old RS-232 dumb terminal from my dad's old computer connected so I could have a shell open w/ an IRC client or my mail while I used Netscape. (I never liked running Pine in an X terminal. It was abnormal and strange. It seemed more natural on the terminal and it was cool to have two screens!)

Yes... I remember that even after having used Linux for a long time when we connected the AT we had the "Unix" experience in a different way.

Reply on news.ycombinator.com