J.Crowcroft@CS.UCL.AC.UK (Jon Crowcroft) (07/05/89)
>We are having trouble reaching a canadian site from some of our >machines - it's only Ultrix 2.2 ones, not earlier (1.2) or later (3.0) >or suns, hps, xeroxes, RTs, pyramids etc...they can all get throough >fine. It turns out that pre ultrix 3 sets TTL from TCP to 15. Our ultrix one machines are on the US side of a local gateway, and 15 hops from canada, so the packets made it with theur last gasp. Our ultrix 2 machines are 16 hops, so never made it. Everything (BSDish) else seems to set TTL to 30 or 60, so is fine (but for how long?). On a more general note, how come we're getting so many hops? - surely the internet is evolving tree structured, so we'd expect log base Gateway/BackBoneFanout(#nets) to set the max hop count (with a bit of random permutation here and there) - and... should local (say subnet) routers decrement TTL? Since they dont participate in global routing, so cannot be part of a global loop? (ok, except they should decrement TTL on time instead of hop, so as to bound TCP segment lifetimes)... jon