Subj : dosxnt To : Paul Williams From : Jasen Betts Date : Sun Sep 29 2002 07:28 am Hi Paul. 27-Sep-02 00:06:08, Paul Williams wrote to Jasen Betts PW> Hi Jasen Betts, hope you are having a nice day PW> JB>Plenty of other explanations exist. I can see no real reason PW> not to JB>simply use sequential numberrs seeded from the date/time PW> when the JB>program was started. unless messages are generated by PW> the node faster JB>than the clock ticks it shouldn't be a problem. PW> That actually was a problem here when I ran the gateways after I'd PW> made a hardware change. Prior to it the system was tossing out 1 PW> msg anywhere from 10 to 90sec apart, however after changing the PW> drive controller to one w/ both a hardware cache and a 286/12 to PW> run things it started tossing msgs on the order of 5-20 msgs/sec. PW> Since systems up stream were *not* using the msgid in dupe PW> checking there were a lot of them being killed off. PW> If I were to run the same setup on a more modern system then I'd PW> be looking at upwards of 40msgs/sec being tossed and needing PW> msgid's created. yeah, once you get up round 40/second clocks aren't a good solution, and CRCs are bad at 100/day the best I can come up with is a "remembered" serial number that's saved to disk after use. so the count can continue from the same point. I didn't implememt that as it's hard to make something like that portable. PW> Well the one upside if I understand the spec right is that a PW> duplicate msgid is only a dupe if it's in the same echo. So even PW> though the below looks like a dupe, it's not as the areatag is PW> different. aas I understand it msgid isn't only used for dupes. PW> It also doesn't leave a mess behind, though I don't see anything PW> in the way of housekeeping in the code. it'd just be what it does when it hits an error - sounds like it just copies bytes until it hits the end of the file. that pktfix program chokes if it hits a char-255 (mistakes it for EOF) PW> That's the way I run it here. When termail exits it runs the PW> archiver to pack up all the pkt's. What I did was to rename the PW> archiver and then write and compile to exe a small batchfile which PW> runs pktfix on the pkt and then runs the archiver w/ the right PW> cmdline options. I do something similar here to strip some of the crap kludges out. PW>>> The kludge that I don't understand is the CHRS one. Why is there PW>>> a PW> PW>>`2' placed after the value? For ex, I use pktfix to add CHRS PW> IBMPC PW>>but the code adds (or did add as i changed it in mine) PW> CHRS IBMPC 2 JB>I think that's correct, dunno why. PW> Dunno either, although I have on occasion seen some that didn't PW> add the 2. Most that didn't were from Z1 which is why I thought it PW> might be a zonal thing. The only ones I've seen withoiuth the 2 were a CHARSET kludge... PW> Which spec is that covered under, do you know? no... none of the files my boss holds mention CHARSET. we should probably move this thread to a different echo, maybe NET_DEV PW>>> Should be interesting to see what you come up w/. PW> JB>yeah... JB>That code chokes on messages with CR LF ^A kludge PW> but although 88 PW> I hadn't noticed anything like that here, looking at the pkt's PW> generated by termail I don't see 0x0D 0x0A 0x01, just a 0x0D 0x01. Yeah, but it puts 0x0d 0x0a in body of the message text, seems that something was barfing on the kludges so they tweaked the code... I don't know why they kept the 0x0d 0x0a in the body though. FWIW something striped them out before it arrived here. -=> Bye <=- --- * Origin: Only God can make random selections. (3:640/531.42) .