After upgrading from RH5.2 to RH6.1, I had intermittent failures with ppp. That is, I could connect every time, but on some connections TCP connections did not work. Control packets (i.e., ping) and DNS always worked. At the suggestion of someone on comp.os.linux.networking, I disabled the "VJ header compression" with the "novj" option in /etc/ppp/options, and this *seems* to have solved the problem (can't be 100% sure, since the problem was intermittent). While this could be a problem at the server end, it seems unlikely, since ppp was working fine with RH5.2, without using "novj". If someone can explain to me how to log the kernel ppp activity, I can perhaps provide better information on what's going wrong (I tried using the pppd "debug" option, but that seems to show only the protocol startup, and doesn't show any IP traffic). Ken McMillan
Quite a few servers get vj compression with options wrong. You may find disabling sack and time stamps has a similar #fixing' effect. For real world use SACK is worth more than VJ except on very reliable links