croft@RUSSELL.STANFORD.EDU.UUCP (03/12/87)
I just updated the files on sumex-aim.stanford.edu:<info-mac> with the KIP 0287 revision. The files changed are: at-kip.shar sources at-gw.srec binary at-telnet.hqx new MacIP MPW Telnet from CMU/Columbia/etc. The file names are all prefixed with 'at-' (appletalk) to differentiate them in the monstrously large <info-mac> directory. If you try a 'dir' command with your FTP, remember to ask only for 'at*.*' files. Other files that havent changed are still available: at-ddt.shar network symbolic debugger at-cap3.shar CAP library from Columbia at-efs.shar external file system (Mac side) Some of these files will also be made available on Mac diskettes through the Kinetics users group; a copy is also mailed out free with the Stanford ACIS/Networking C-based MacIP package. Here is a list of the revisions in KIP since the last release: ---- kip/gw.c/iematch(), abmatch(); unsigned compares failing The range checks for ARP matching on the IP dynamic addresses are failing sometimes due to unsigned compare wraparound. New code rewrites these functions to avoid unsigned compares. Since Stanford has a class A address, we didnt uncover this until class B and C sites started using the code. [reported by mckay@ee.ecn.purdue.edu, hemphill@nrl-aic, etc.] kip/gw.c,gw2.c/; hop counts in ddp header The DDP hop count is now incremented and masked off properly when looking at the ddp.length field. [reported by kre (Robert Elz)] kip/rtmp.c/artinput(); allow arbitrary format broadcast addresses The old code made assumptions about the route table entry for the directly attached ether cable. The new code allows this to be set instead from the atalkad. [reported by mckay@ee.ecn.purdue.edu, dk1z#@andrew.cmu.edu, etc.] kip/gw.c/routeddp(); route failure with interbridge An extraneous routing loopback check was discarding appletalk packets with src&dst on the same cable. Check removed. [reported by gcc-milo!brad] kip/etc/atalkad.c; 4.3 BSD function name conflict Rename the atalkad.c function getshort() to be getashort(). 4.3 domain name server library already has a function called getshort(). kip/etc/atalkad.c; zero zone name field Since zone name field length may change upon a reread of atalkatab, that field should be zeroed in the structure passed to the kboxes. [reported by gcc-milo!brad] kip/etc/atalkad.c; (month/date) in atalkalog January is being reported as month number zero(!) cap/samples/papif.c assumes special lpd Papif assumes that lpd has been modified to pass the printer name to the filter with the '-p' switch. To avoid this dependency a trivial shell file can be setup that execs in papif with the proper -p switch along with the previous arguments. kip/etc/broadcast New documentation that better explains IP broadcasting issues.
cck%cunixc@COLUMBIA.EDU (Charlie C. Kim) (03/16/87)
The newest MacIP Telnet (MPW version) that Bill Croft just placed into the info-mac archives is substantially more robust than the previous copy. The previous copy had a major bug in the code that dealt with the new KIP software and it was just plain old luck that it ever worked. Outstanding bugs/features that I can remember are: o You cannot switch working directories using "Commands" Sendfile command. Attempting to do so will cause telnet to bomb. Anyone who can figure out how to fix this or even just what the problem is, please let everyone know! o The terminal emulation code isn't as robust as it should be. For example, updates don't quite happen correctly when the telnet window is the active window, but not the front window. Some of the screen commands don't do range checking either... Also, no provisions for remapping the "backspace" (on the plus)/"delete" (on the SE) to backspace vs. rubout (this always ends up driving me up the wall). Anyone who wants to fix these things... o Some problems may occur with using desk accessories... I think this is due to stack space problems, but am not sure. o The new TFtp menu is nonstandard and more than a little silly, but it does work. If anyone has a clean way to handle the menu, go ahead... (dialog box maybe?) The intent is to provide the server functionality that the old tftp program did; however, all the user tftp support is still missing. o The "Send" "IP Address" has been reenabled; I suppose it was dropped to allow telnet to run on a 128K Mac in the past. It is particularly useful with the dynamic IP address assignment. o Won't run with TFTP in less than O(200KB). Should run on 512. I would be surprised if it didn't run on a 512E. However, hasn't been tested on anything except a Plus or SE. May not run at all on a 128K Mac. o You cannot switch "working directories" for file transfers - it is the directory that telnet was launched from; however, you should be able to specify the fully qualified name as "vol:directory:file" (it is MPW style where the prefix "::" means one up and the prefix ":" means current directory). o This version of telnet still uses the old IEN116 name service, not the domain name service; however, I think it actually works correctly now (there was an alignment problem before). o The copy that went out has the maximum number of TFTP connections allowed set to two (e.g. you can be doing two tftp xfers at a time). No particular reason except that we were playing with it to see if that code worked or not. Shouldn't cause any problems. o Aborted file transfers (downloads) sometimes result in files remaining open and thus locked. Basically, while quite a few bugs remain, you should find it as useable and more robust than the "Lisa Pascal" version available before the advent of KIP1086. Charlie C. Kim User Services Columbia University