STJOHNS@SRI-NIC.ARPA (Michael St. Johns) (12/30/87)
I'm in the process of trying to get a micro vax II running Ultrix up and operational on an ether net behind a cisco gateway attached to the Arpanet. I am experiencing some problems. When I try to transfer (FTP or TELNET) more data than will fill a 512 byte TCP segment, the connection hangs on me. This transfer is either by doing something like "more /etc/termcap" or starting an FTP transgfer of a file larger than 512 bytes . Here's the kicker... this only happens on transfers outbound from the micro vax, inbound transfers seems to work fine. Transfers on the same ethernet seem to work fine in either direction and with any amount of data. Raw PINGs seems to work fine for ANY size ping packet. In either direction. We've swapped the ether interface card in the gateway with no effect. Has anyone else seen this type of behavior? This is an ULTRIX 2.0 system (binary version *sigh*) so I haven't been able to peek real closely at the innards. Help! Mike -------
StJohns@SRI-NIC.ARPA (Mike StJohns) (12/30/87)
And the answer is... ifconfig qe0 -trailers My thanks to all (at last count 8) the people who submitted this answer, but I regret I can't award duplicate prizes *grin*. And my boss doesn't understand why I like the net so much. Mike
medin@AMES-TITAN.ARPA (Milo S. Medin, NASA ARC Code ED) (12/30/87)
I thought Ultrix 2.0 did what 4.3 does, that is negotiate during the ARP for trailer use? That way you can use trailers to hosts who support it, and not use them to hosts who don't... Oh well, maybe next time. Vendors should note that 4.2 networking code just isn't good enough anymore... Milo
louie@TRANTOR.UMD.EDU ("Louis A. Mamakos") (12/30/87)
Gee, another chance to bash Ultrix networking code :-> The networking code isn't as bad as that Milo; they've moved up to a 4.3 alpha or beta release before the tailer negotiation went in. I'd be real suprised if the release version of Ultrix 2.2 has the trailer negotiation in it either. I guess after putting in DECnet, LAT, and other stuff most UNIX customers don't want, they ran out of time and didn't get a chance to port the release 4.3 stuff. It's not that Ultrix is crummy or that the folks in the Ultrix Engineering group are stupid. They're not. They've done some real neat stuff. It's just not stuff that I (the customer) really wants. Oh well, maybe in 2.4 we'll even get Van's new TCP (though I doubt it). louie
medin@AMES-TITAN.ARPA (Milo S. Medin, NASA ARC Code ED) (12/31/87)
Hmmm. That's pretty amusing. The Wollongong TCP/IP code we run here at Ames (v 3.1) is all 4.3 BSD based kernel code (thanks mostly to Jerry Scott), and they've even found 4.3 bugs and fixed them! This means that a VAX running VMS with the TWG code has better TCP/IP support than the same hardware running Ultrix. Gads, I never thought I'd see the day... Of course, 4.3 BSD (Unix) itself runs really well... Milo
bassen@IFI.UIO.NO (Tor Sverre Lande) (01/05/88)
Well, there is a couple of more hints using TCP/IP on Ultrix. 1. If you also are running DECNET do ALL the DECNET-stuff before "ifconfig" in rc.local. The DECNET-software changes your ehternet-adress so if you do "ifconfig" before "ncp ...." your TCP/IP will never work. Did you ever think hardware ethernet-addresses should be unique?!? 2. If you are using subnets and have a gateway taking max. 576 bytes packets between your subnets there is an undocumented kernel-option NETCONSERVE setting max packetsize to 576 to all non-local subnets. This is the same option as SUBNETSARELOCAL in 4.3bsd. -Bassen