Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!uunet!tut.cis.ohio-state.edu!hsdndev!husc6!harvarda.harvard.edu!conrad From: conrad@harvarda.harvard.edu (Conrad C. Nobili) Newsgroups: comp.sys.mac.comm Subject: Re: MacTCP problem Message-ID: <4890@husc6.harvard.edu> Date: 2 Dec 90 20:49:56 GMT Sender: news@husc6.harvard.edu Organization: Harvard University Lines: 74 References:<1990Nov30.205337.8345@ux1.cso.uiuc.edu> <693@nih-csl.nih.gov> In article brad@andros.cayman.com (Brad Parker) writes: > > In article <693@nih-csl.nih.gov> ken@helix.nih.gov (Ken Weeks) writes: > > I've been using NCSA Telnet to talk to my VAX, which is running a Telnet > server. I recently "converted" from the embedded TCP driver version to > the Mac TCP version. No problems at first...but as of late, I've noticed a > severe, repeatable "hang" when I use the VMS "TYPE" commands, or list a > long directory. Just that session hangs - I can log in again and kill it > > I think the "mss" or maximum segment size of the TCP you are using is > too large for your gateway. Under NCSA you can set it (in the > config.tel file) to a small number like 512 which will not require > that you gateway fragment the packets. Under MacTCP, is it not settable, and > I think MacTCP 1.0 has a bug which causes it to guess wrong; (or at least > guess a number which causes you to need a gateway which will fragment > properly) I would be interested in some clarification of this possible MacTCP 1.0 bug. Brad, did you not notice that Ken said he was using MacTCP 1.0.1? Does your use of the "1.0" version number mean that you think 1.0.1 fixes the bug? Or does it mean that it won't be fixed until version "2.0"? (Here I am making the assumption that 1.0.1 is still the current version. Do you developers have a newer one I don't know about?) > I'll go out on a limb and guess that your mac is on localtalk and you're > using a gateway to get to the ethernet. If the gateway is a fastpath, it > may not support fragmentation of IP packets too big to be encapsulated > in a single localtalk packet (~600 bytes); If the gateway is a gatorbox, > you're software may be out of date as we used to have this problem > but fixed it. Also, I am a bit puzzled by your suggestion that the "mss" is too large for the gateway. Unless I am confused about how the config.tel file works, your suggestion to set the maximum segment size higher there doesn't make sense in light of your "out on a limb" guess about Ken's setup. I.e., if indeed Ken does have a Mac on LocalTalk and is using a gateway to get to the ethernet, does he not HAVE to have "hardware=AppleTalk" in his config.tel file for non-MacTCP versions of NCSA Telnet to work at all? And if so, does he not HAVE to have "maxseg=512" in his config.tel file already for non-MacTCP versions of NCSA Telnet to work at all? Can Brad or others clear me up on these things? The reason I am curious about all this detail is that Ken's problem has some similarities to the one I posted yesterday. ("Liaison: TCP/IP services & dial-in access problems" <<1789>>) I know that mostly there are differences, but I too have a "severe repeatable 'hang'" after having been logged on ok for some time (in my case it seems to be a TIME-dependent thing after five minutes), and, having a Mac on LocalTalk and a GatorBox as the gateway, I was interested in Brad's ideas on the subject. (By the way, Brad, my GatorShare 1.5 software is not current (as of last week, I think 8-) ), but current enough not to be "out of date" as you meant it to Ken, if I read the GatorAid 1.5 handbook correctly, so that's not MY problem....) I apologize if this should have been e-mailed to Brad rather than posted to the net. Can someone flame me via e-mail and clue me in if that's the case? Thanks. C o n r a d C . N o b i l i Harvard University | Internet: conrad@harvarda.harvard.edu Office for Information Technology | BITNET: CONRAD AT HARVARDA Information Services | voice: (617) 495-8554 Technical and User Services | fax: (617) 495-0715 < < < D i s c l a i m e r > > >