MOO-cows Mailing List Archive


chat room

MOO is VR.  It emulates, to an approximation, what would happen in RL.
Its approximation of what would happen with 2000 people speaking at
once in one room is really quite close to what would happen in RL.
That is, unintelligibility.  Imagine yourself in a nice large theater
discussing a play you just saw with literally everyone in the audience
speaking at the same time.

Ever watch an electronic town meeting (or a real one)?  You get a lot of
people together.  How many get to speak?  How long would the meeting take
if they all spoke?  VR or RL, it's pretty much the same...

I concur with the sorts of things that others have already posted,
ESPECIALLY about the O(N^2) problem.  And gagging as a basis for figuring
out who should be speaking to whom is not really going to be a very
efficient mechanism.  Gagging works when gaglists are short--it's not clear
it works as well when your default is to mostly gag everyone and gaglists
get enormous or when the number of people with noticeably long gaglists is 
also enormous.

Why not take advantage of the one structural element that MOO offers which
naturally fights lag: locality of reference.  Make it easy for people to 
speak locally and hard to speak globally: scatter people into different rooms.

If you must have large groups, an auditorium-style thing could be
constructed with only O(N) badness, as the speaker speaks to N people
but they don't attempt to respond.  Give one person a room in which to
speak or give that person a microphone and turn off the ability of the
others to reply.  In fact, once you do this, your audience can be
distributed into small rooms that have "TV screens" that see the main
speaker and that allow small, localized chat sessions among small numbers
of people...

Some data points...

 - I also have seen 30-something people in a MOO room at once (at Dred's
   at LambdaMOO) and it's pretty close to unbearable pragmatically--the 
   lag is pretty bad and the spam is awful.

 - At July 4, there was a fireworks show at LambdaMOO which had about 90
   people trying to watch (from different rooms in a broadcast style).
   For some it finished inside of 15 mins.  Others reported begin so
   lagged they didn't see the end for more than 2 hours after that.
   Some of this, it was hypothesized afterwards, could have been
   solved by taking advantage of the knowledge that the people had
   subscribed to the fireworks and hence didn't need a fresh
   gag-filter applied on every bang and boom.  In retrospect, too
   little work was done trying to tune the efficiency to accomodate so
   large a crowd.  Probably it should have gotten wiz support for
   :notify() instead of using :tell().  That would have cut down by a
   large constant factor if not more.  Of course, for your purposes, you
   must keep in mind that you don't get gagging with :notify() ...

 - At LambdaMOO, even with people in different rooms not all talking at once,
   lag tends to become noticeable when there are 200+ people doing things at
   once.  Perhaps some of those people are doing more than "just talking".
   And perhaps there are background processes doing things like mail 
   notification that are taking up time.  But in any case, that's well short
   of a thousand, and they aren't even all in one room.  Many other MOOs I
   know lag down at 30-50 people (again, not in same room).


Home | Subject Index | Thread Index