Author: Ratko V Tomic
Date: 19:59:22 10/04/99
Go up one level in this thread
For a portable protocol a strict separation ought to be made between the protocol language (covering the game managment) and the means of transmission. With such separation, the basic information packets would be well defined and portable to whatever means of communication is available between the machines. One would need only the basic driver for a given type of link which would be separate from the chess program. The chess program would only communicate to a named device driver (on most systems such driver is opened just like a file, then one writes or reads from it) using the language part of the specification and would never worry about opening sockets or COM channels or whatever else the driver on a given platform needs to transmit the data. In case of TCP/IP, the DOS versions of the protocol are not widespread or of a good quality. Those I saw few years ago required a large and aggressive TSR to be in memory. So that wouldn't work if a program needs 500+ K of DOS main memory (that's different from hash tables in extended memory) to run. Basically, carrying the overhead of TCP/IP stack under DOS for just transmitting the few bytes per second that come in every minute or more is an overkill. Of course, one may abandon DOS versions of the programs. While this is probably Ok for end user product, I think in a competition, one may wish to squeeze as much out of the machine as possible, and machine without Windows, running plain DOS will give you many magabytes more of RAM as well as at least 5-8 percent greater CPU speed (Windows has lots of background activity, whether your program asks it for it or not).
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.