When eCS 2.0 boots up everything runs very smoothly. I can get on the Internet and things are fast, life is good.
I run three programs on the Internet most of the time:
1. Firefox - ver 3.5.3
2. FTP Browser - 1.71
3. GTIRC - Ver - 3.00
After 30 minutes (to the second), Anything that is on the Internet loses connection and I go no where. The very weird thing is that GTIRC still works!?!? I still can have and receive conversations will people for hours. The only way I can get Firefox and FTP Browser working again is to reboot, then the clock starts again until 30 min elapses, then I have to reboot again if I want to surf.
I thought of a few things.
(1) IRC uses TCP like most other things you are using (as opposed to UDP, etc) - I don't see how it specifically would be a problem.
(1a) After 30 minutes, when the problem occurs, if you close GTIRC and restart it, will it work? If not, then it too isn't immune, and simply is working in your previous tests because it still
has it's sockets open and active. That would mean that the programs are having problems
creating new sockets.
(2) I have had similar problems on Windows running Firefox. I never did figure out why, but since I upgraded Firefox, I have not had that issue anymore. Perhaps there is something buggy in the 3.5.x release(s) that causes this? I was running a 3.5.x (not sure which .x) version on Windows when this problem would crop up. It wasn't like clockwork, but it happened every time I surfed for a bit.
(3) Do you have NetBIOS over TCP/IP enabled? If so (and you don't need it), perhaps try disabling it and see what happens. I *DO* know, without a doubt, that my OS/2 machines, with or without that enabled, spend a lot of time ignoring all sorts of requests from my one Windows box. Not sure how that can affect an OS/2 or eComstation box, but perhaps it is causing an issue.
(4) You can test for certain things that may be causing problems (for instance, something causing the scenario in #1 above) by (from a command prompt) running:
netstat -m
- Check (usually) the last two entries for anything greater than a single digit number
(mbufs: buffers used/allocated by your card and the networking subsystem to buffer/handle I/O)
- dangerous to change these values unless you like trap errors and such, but they would be changed
in the socketsk.sys entry in config.sys
netstat -s
- Check to see how many entries... if you have a massive list, something is wrong
(sockets connected: for instance, well... anything that uses the Internet or a TCP/UDP based intra/internet)
netstat -l
- Check to see how many entries... if you have a massive list, something is wrong
(open sockets in listen mode: for instance server daemons, IRC clients, etc)
(5) You can check your TCPIP config (tcpconfig2 from the commandline in older versions of eCS and OS/2 - or the TCPIP Config icon in most versions) and ensure (Network -> Advanced Options):
- (Page 1 on original style notebook)
Broadcast: blank (these are configured under routing)
Destination: blank (these are configured under routing)
Metric Count: 1 (usually - 1 will work for 99% of every scenario)
Max. Transm. Unit: 1500 (unless token ring or you know what you are doing)
- values greater than 1500 will cause all sorts of weird issues with various NIC drivers/cards
- (Page 2 on original style notebook)
I generally have nothing checked
- Changing things like icmp redirects and such can cause interesting problems on certain setups/networks
Best,
Robert