mosemann@sardion.unl.edu (Russell Mosemann) (03/27/91)
Weeks and weeks ago, people were discussing the problem MacTCP has with determining the MSS (maximum segment size). I read the messages but never wrote anything down or saved it because we didn't have that problem before. A few weeks ago MacPOP (from trident.arc.nasa.gov) started complaining that there was an error connecting to the POP server. I could telnet to the port and issue commands by hand with no problem. Using a network sniffer, I saw the domain name request get resolved and the Mac sent out 4 or 5 TCP packets to the POP host, but the host didn't respond. I dug into the packets and found that it was requesting a window size of *10241*. Could someone mail the fix for this (I think it was done by Steve Dorner)? What was the explanation for this? I think it had something to do with a device on the network advertising the wrong segment size. If so, is there any fix for that device besides turning it off? The other weird thing is that only the MacPOP software seems to be affected. I can telnet with TCP/Connect II and NCSA Telnet just fine. Thanks for the help. -- Russell Mosemann Internet: mosemann@unl.edu Network Analyst Bitnet: mosemann@unlvax1 Computing Resource Center UUCP: ..!uunet!hoss.unl.edu!mosemann University of Nebraska - Lincoln Voice: (402) 472-5930 Fax: 472-5280
mosemann@sardion.unl.edu (Russell Mosemann) (03/27/91)
In <1991Mar26.213034.8544@hoss.unl.edu> mosemann@sardion.unl.edu (Russell Mosemann) writes: > Weeks and weeks ago, people were discussing the problem MacTCP has >with determining the MSS (maximum segment size). I read the messages >but never wrote anything down or saved it because we didn't have that >problem before. [...] > I dug into the packets and found that it was requesting a window size >of *10241*. Could someone mail the fix for this (I think it was done by >Steve Dorner)? [...] OK. I got the fix, but it didn't solve the problem. Now for a little more information. I'm using MacTCP 1.0, MacPOP 1.5.7, Systems 6.0.5 and 6.0.7, and Ethernet'ed and LocalTalk'ed Macs (SE/30, SE, and Classic). MacPOP tries to establish the tcp connection. I see the 4 or so packets go to the POP host, but there is no return response. Our sniffer reports: Ethernet: ( router -> pop host) type: IP(0x0800) Internet: Mac IP -> pop host IP hl: 6 ver: 4 tos: 0 len: 48 id: 0x0009 fragoff: 0 flags: 00 ttl: 58 prot: TCP(6) xsum: 0x6a75 bsec resv 4: TCP: 10593 -> POP-2(109) seq: 5d60a560 ack: ---- win: 11648 hl: 6 xsum: 0x3a3a urg: 0 flags: <SYN> mss: 576 Only the Ethernet and IP addresses have been changed. This is from a Mac Classic going through a FastPath 4 running version 8.x of KSTAR. I can telnet to the port and issue commands by hand with no problem. The mss doesn't seem to be the problem that I thought it was. The window size seems a little strange, but maybe it's OK, too. The thing I have never seen in a TCP packet before is the "bsec resv 4:". Does anything look wrong in this packet? Thanks. -- Russell Mosemann Internet: mosemann@unl.edu Network Analyst Bitnet: mosemann@unlvax1 Computing Resource Center UUCP: ..!uunet!hoss.unl.edu!mosemann University of Nebraska - Lincoln Voice: (402) 472-5930 Fax: 472-5280
mosemann@sardion.unl.edu (Russell Mosemann) (03/27/91)
In <1991Mar26.213034.8544@hoss.unl.edu> mosemann@sardion.unl.edu (Russell Mosemann) writes: > Weeks and weeks ago, people were discussing the problem MacTCP has Arg! I forgot to mention that the pop host is a Sun (we've tried a Sparc 1 and a Sparc 2) running SunOS 4.1.1. MacPOP was working a month or two ago. One thing that changed in that time was that we moved from Sun0S 4.0.3 to 4.1 to 4.1.1. -- Russell Mosemann Internet: mosemann@unl.edu Network Analyst Bitnet: mosemann@unlvax1 Computing Resource Center UUCP: ..!uunet!hoss.unl.edu!mosemann University of Nebraska - Lincoln Voice: (402) 472-5930 Fax: 472-5280
JSIMPSON@MIAMIU.BITNET (Joe Simpson) (03/30/91)
This is a known problem . MacTCP version 1.0.1 fixed it.