Bandwidth consumed by (non-webbed) MUDs is very low. Lets look at a few figures taken from postings to the cbnvee mailing list by S. Alexander Jacobson and Lydia Leong on Jun2 27 1995.
A t1 network connection (the usual setup for most universities) has has a bandwidth of 1.5 mbps.
``If we divide that bandwith among 5,000 computers, we get an effective bandwidth per computer of 300 bps. You might argue that 300 bps is awefully slow, but consider that in the case of MUDS/MOOS we are talking text here. There is no way that the partipants are generating even close to 300 bps. If everyone at the school could type 90 words per minute (ALL DAY LONG), that would amount to only 12 cps or 96 bps. Even if text was coming from outside the university at the same speed, that would amount to only 192 bps, still easily within the bandwidth of the single t1'' (S.A. Jcobson, above posting)
``To give you some hard numbers: five minutes ago, I grabbed the logfile from Patternfall MUSH, a small-to-medium sized TinyMUSH, which gets about the same sort of traffic places like BayMOO get. Over the course of slightly less than eighteen days of uptime: 108,126 commands were entered, totalling 4,471,099 bytes of input. 55,292,081 bytes of output were sent back in response. This data is sampled over 2901 connections to the game. That tells us that, on average: 37.27 commands are entered per session, totalling 1,541 bytes of input, and resulting in 19,060 bytes of output. The average user thus enters 1.5K of text per session, and receives 18.6K of text back.'' (Lydia Leong, above posting).
Roughly, this means that active ``talking-oriented'' MUD sessions takes about as much banchwidth as the transfer of a single text-based HTML page. Sessions where people retrieve a lot of help text or other structured documentation will amoung roughly to the retrieval of a single multi-media HTML page.