andyrose@uunet.UU.NET (Andy Rose) (09/27/90)
An excellent start to modeling virtual worlds must be credited to the effervescent nocturnal gladiators. telnet 128.2.250.158 2323 will log you into a thing called Islandia, which is a MUD, a multi user dungeon. Say create <yourname> <a password> to get going. When you appear in the bus terminal say 'out' and you will go to the old oak tree. Here you will "see other players". This is in the form of a text line like "Aiko is here." Players are apparently allowed to add features to the environment which others interact with. A facility for running an automatic player (a robot) is also apparently provided. I have no experience in this realm, so don't ask me... SO * a description of an area appears (after you cross a threshold like a door) "In the forest with an east-west path covered with mulch." Left brain blasts a perspective view of typical old growth forest (pines, etc.) with a horizontal stripe of some width textured with bumpy shades of brown. How far are we really from having a context sensitive language parser which could make these inferences (I contend not far). Viola, local rendering of complex visual scene from less than 80 byte description. * You idle, getting used to the feedback delay of your new virtual light saber driven by your power glove. WHEN "Donald Knuth walks in from the east. He does not notice you." Here the local renderer peruses the data base for a bit map of Knuth, or even a 3D topology of his head (Failure Analysis, Siggraph '90 digitized many folks with some kind of laser contraption which rotated around), and bobs it along, or if the compute power is there adds a body. Note that the renderer knows not to have the eyes appear to look "at you". * You, wishing to attract Don's attention to clear up some confusion about the Approximate Behavior of Caron's Polyphase Merge Sort (p.281 V.3), wave your saber about. Somewhere Don sees the message "you can see <yourname> waving his saber" (Don has no workstation to render on so he is using strictly text." Don's "deck" knows where Don can see, so... * Don types "say hi". You see "Don says hi." You voice your concern about sorting and merging. Don clears everything up for you and sends you e-mail with some old algol code. This approach relies on local servers to recognize protocal from a client. The client would have access to a database of room descriptions, faces, objects, etc. The trick here is distributed clients, multiple clients, shared data, bandwidth, and protocal definition (to name a few). The cleverer the server, the less load on the client. I think that the MUD approach clearly shows the way for multiuser databases but has a restriction in data sharing. Each client must maintain its own database (which is pragmatically reasonable now). Note that interaction between you and Knuth is simply textual message passing and is not as bandwidth intensive as say a laser sword battle between 40 users. (Checking for polygon intersections...) The real kickdown will be in a distributed data environment. As supercomputer speeds become available on desktops you can imagine the "supercomputers" becoming massive file servers with all kinds of virtual worlds (databases) on-line. Of course, there is more, but for now this is enough. "The faster we go, the rounder we get." andy@cornellf.tn.cornell.edu -- Andrew Newkirk Rose '91 Department of Visualization Cornell National Supercomputing Facility / Theory Center 632 Engineering and Theory Building