Re: Misc notes/thoughts.

[Drew Eckhardt <>]

In message <>, krisb@seat
tleu.edu writes:
>
>On Mon, 11 Dec 1995, GESTIONNAIRE DU Casino wrote:
>
>> On Mon, 11 Dec 1995, Kristofer M. Bosland wrote:
>> 
>> 
>> So when you leave a map, you flush all the data for it, and load all the 
>> new one... Why not do a bit a caching ? 
>> Usually maps won't change a lot.... So the client could keep the maps in 
>> memory for a certain amount of time, say 10 minutes, and when the player 
>> enters a map, the client either asks for the whole map, if it doesn't 
>> have it in memory, or aks for a diff since time x.
>> But then, the server needs to remember :-) 
>> What is the important factor ? optimising the client/server 
>> communication, or the memory requirement of both client and server ?
>> 
>> ---
>> Casino
>> 
>
>	Well, first I think that this is less important than agreeing on
>the "Architecture" of the communiation.  By this, I mean to address
>Orthoganality and Generality, and to enable more possible clients (i.e.
>Tcl/Tk can't handle 0x00 generally, so text based communication is better
>than some binary format).  The Servers memory requirements will probably
>not change much, but may go down (I wouldn't expect the server to remember
>what the clients contain).  As far as BW Vs. Client Memory Requirements, I
>think that we should lean towards more mem, less BW.  The choke point for
>many people is their 1 byte/sec comm link, not their 100MHz CPU or 8MB
>Memory.

Latency is the other big issue - it takes me 500ms and 30,000 miles 
to make the round trip between my house and office three miles down the 
road.