PCjr Telnet BBS Test
Michael B. Brutman
mbbrutman-cctalk at brutman.com
Sat Jun 20 18:16:53 CDT 2009
Andrew Burton wrote:
> Well, you did want us to test it out and try and break it, lol.
> Just glad I didn't bring it down completely.
> Is there a way to adjust the retransmit time depending on user location, or is that impossible / too complicated?
> Or will it be more friendly to everyone once it is increased a little?
> Andrew B
> aliensrcooluk at yahoo.co.uk
Cosam and I were talking about it and the right neurons fired in my brain.
Back in December I made a change to the way I measure elapsed time. The
change gave me like a 30% performance improvement because my original
code was so bad, and the new code was counting timer ticks which is
pretty easy. When I made the change I changed all of my units of time
measurement from 100ths of seconds to milliseconds.
Except in the configuration file that defines the timeouts for the BBS. :-)
The result is that I'm retransmitting packets if I don't get an
acknowledgment in 0.4 seconds, not the 4 seconds that I intended. Which
explains why your particular modem connection was driving the
retransmits so hard. Most broad-band users don't have that problem.
It's not bad enough to stop and restart the BBS, but it definitely is
exercising that path in the code pretty hard.
More information about the cctalk