hthompso%eusip.edinburgh.ac.uk@NSFNET-RELAY.AC.UK (Henry Thompson) (01/16/90)
One of the reasons for investigating PUP-FS (see previous msg) is that using TCP-FTP is really losing: 5 HOURS to sysin from an otherwise unloaded Sun file server, and erratic hanging when trying to sysin from another 1186 on the same net, running TCPFTP.SERVER. Since normal lisp-based TCP file transfers seem quite brisk, I conclude it's the SystemTools code which is at fault. Anybody any ideas/patches/similar experiences? Henry Thompson, Department of Artificial Intelligence, University of Edinburgh 80 South Bridge, Edinburgh EH1 1HN, Scotland Telephone: 031 225-7774 ARPA: hthompso%ed.eusip@nsfnet-relay.ac.uk JANET: hthompso@uk.ac.ed.eusip UUCP: ...!uunet!mcvax!ukc!eusip!hthompso
postel@VENERA.ISI.EDU (01/17/90)
Maybe your TCP need to be brought up to speed with the latest improvements from Berkeley 4.3 and Karn's algorithm and Jacobson's algorithm. Please see RFC-1122 for numerous details on improving TCP implementations. --jon.
hthompso%cstr.edinburgh.ac.uk@NSFNET-RELAY.AC.UK (Henry Thompson) (03/15/90)
Turns out the problem was not with the address at all, but with IEEE 802.3 compatibility switch -- Medley sysouts/microcode default to old, non-compatible encapsulation types, but System Tool defaults to new, compatible types. Changing either one enables successful communication, and Sysin times not much worse than from XNS file server. ht