Resource Contention: CALL, DISPLAY OR MODIFY ?
One of my friends mentioned difficult parking at the mall. The parking was bad enough that she did more shopping than originally planned. As a computer science type, this sort of thing brings to mind discussions of resource contention. Would the mall make more money by having more parking? There's very little change in shopper behavior as long as the lot still has a few percent of its spaces free; people will do their nominal amount of shopping. As the number of free spaces approaches zero, there's suddenly a big problem. Once somebody manages to get a space, they're inclined to use it for longer, exacerbating the problem. This is rational behavior for the individual, but problematic for the system as a whole.
It brought back memories of an even worse situation in college. The University of Rochester once represented the world's largest installation of ROLM-brand phones. These were some fancy phones . . . the entire campus's phone system was digital and everyone had voicemail. It was a nuisance that people couldn't bring in their own phones, but there was a critical feature that made up for it. The campus hadn't been wired with Ethernet, you see, and the phones had a digital modem in them which allowed 19,200 baud text connections to the various servers on campus. One had only to plug in a power brick and run a serial cable from the computer to the back of the phone and fire up a terminal emulator. The phone would say "CALL, DISPLAY OR MODIFY ?" and users would respond with something like "c uhura", which would connect to Uhura, the mail server. It wasn't useful for just email . . . some classes received important announcements via newsgroups. If you were taking certain physics classes, it was even possible to submit homework answers online, and the server would let you know if each answer was right or wrong, allowing students to try again until they got it right.
The campus was too large to be one monolithic ROLM network, so at some point they subdivided it and ran some finite number of lines between the two halves. If you were on the right end, it was still possible to connect directly. To the best of my knowledge, this rarely presented an obstacle unless a server was critically overloaded. I, of course, had the misfortune of being on the wrong end of the schism. We couldn't connect directly to any servers. We had to "c net" first. Once that connected us to the main network, we could "c uhura" and go about our merry way.
The problem was the finite number of lines over which one could "c net". I think there were 120 of them, which needed to suffice for approximately 1,000 dorm rooms, at least 150 of which were doubles. Resource contention was insanely high at peak times. The phones would put you in a queue if you tried to c net and the system was at capacity, but the queue was also finite. Some times it wasn't even possible to get in line! Once per minute, you'd be notified of your position in the queue. Once you made it to the head and a line became available, the phone would beep once. You had a few seconds to type something, or the system would boot you out to make room for someone else. When any of us heard that beep, we'd go running to try to catch it, even if it wasn't ours. Once someone did get in, the neighborly thing to do was to offer anyone else within earshot a chance to check their mail or submit some homework answers. As you can imagine, a single connection would sometimes last for hours as everyone had a turn. Had there not been so much difficulty, a typical connection might only be long enough to confirm that one didn't have any new mail.
The situation with the modems was so dire, we'd resort to trickery to jump the line. After a c net connection was established, the phone would report to which five-digit extension it had connected. It was possible to try connecting to those directly, which had the potential to bypass the load-balancing of c net. Those of us who were already numerically-minded had little difficulty memorizing a few of them for just such a situation. Pity the person who'd been in queue for an hour who suddenly found themselves at the back of the line because someone else had managed to connect to the right number at the right time.
To get a sense of how frustrating all of this was, imagine a parking lot. Once the lot fills up, nobody would leave a space without calling their friends and saying "I'm leaving a parking space at the mall . . . should I wait to leave until you can drive over and take it?" And sometimes, just when you'd thought a space had actually opened up in front of you, a car would drop into it from the sky.
I probably don't need to tell you that my dorm was wired for Ethernet the Summer I graduated.
It brought back memories of an even worse situation in college. The University of Rochester once represented the world's largest installation of ROLM-brand phones. These were some fancy phones . . . the entire campus's phone system was digital and everyone had voicemail. It was a nuisance that people couldn't bring in their own phones, but there was a critical feature that made up for it. The campus hadn't been wired with Ethernet, you see, and the phones had a digital modem in them which allowed 19,200 baud text connections to the various servers on campus. One had only to plug in a power brick and run a serial cable from the computer to the back of the phone and fire up a terminal emulator. The phone would say "CALL, DISPLAY OR MODIFY ?" and users would respond with something like "c uhura", which would connect to Uhura, the mail server. It wasn't useful for just email . . . some classes received important announcements via newsgroups. If you were taking certain physics classes, it was even possible to submit homework answers online, and the server would let you know if each answer was right or wrong, allowing students to try again until they got it right.
The campus was too large to be one monolithic ROLM network, so at some point they subdivided it and ran some finite number of lines between the two halves. If you were on the right end, it was still possible to connect directly. To the best of my knowledge, this rarely presented an obstacle unless a server was critically overloaded. I, of course, had the misfortune of being on the wrong end of the schism. We couldn't connect directly to any servers. We had to "c net" first. Once that connected us to the main network, we could "c uhura" and go about our merry way.
The problem was the finite number of lines over which one could "c net". I think there were 120 of them, which needed to suffice for approximately 1,000 dorm rooms, at least 150 of which were doubles. Resource contention was insanely high at peak times. The phones would put you in a queue if you tried to c net and the system was at capacity, but the queue was also finite. Some times it wasn't even possible to get in line! Once per minute, you'd be notified of your position in the queue. Once you made it to the head and a line became available, the phone would beep once. You had a few seconds to type something, or the system would boot you out to make room for someone else. When any of us heard that beep, we'd go running to try to catch it, even if it wasn't ours. Once someone did get in, the neighborly thing to do was to offer anyone else within earshot a chance to check their mail or submit some homework answers. As you can imagine, a single connection would sometimes last for hours as everyone had a turn. Had there not been so much difficulty, a typical connection might only be long enough to confirm that one didn't have any new mail.
The situation with the modems was so dire, we'd resort to trickery to jump the line. After a c net connection was established, the phone would report to which five-digit extension it had connected. It was possible to try connecting to those directly, which had the potential to bypass the load-balancing of c net. Those of us who were already numerically-minded had little difficulty memorizing a few of them for just such a situation. Pity the person who'd been in queue for an hour who suddenly found themselves at the back of the line because someone else had managed to connect to the right number at the right time.
To get a sense of how frustrating all of this was, imagine a parking lot. Once the lot fills up, nobody would leave a space without calling their friends and saying "I'm leaving a parking space at the mall . . . should I wait to leave until you can drive over and take it?" And sometimes, just when you'd thought a space had actually opened up in front of you, a car would drop into it from the sky.
I probably don't need to tell you that my dorm was wired for Ethernet the Summer I graduated.