falken@caen.engin.umich.edu (David R Falkenburg) (01/16/89)
Here at University of Michigan, we're attempting to establish a campus-wide (and eventually maybe encompasing the Merit Network) AppleTalk internet. We're interested in hearing from anybody out there who mmight have attempted to build a large internet (>= 20 - 50 Gateways) Please respond via electronic mail, & I'll get back with you. -dave falkenburg -- Dave Falkenburg @ University of Michigan Computer Aided Engineering Network ARPA: falken@caen.engin.umich.edu UUCP: umix!caen.engin.umich.edu!falken
asher@UHURA.CC.ROCHESTER.EDU (Samuel Asher) (01/26/89)
Interesting-- John Castler from Syracuse University and I have requests from our respective superiors to do what it takes to tie U of Rochester and SU appletalks together-- at least have remote printing, if not total zone transparency. Things have been somewhat on hold for me because of other responsibilities, but here is what I have so far: Problems to overcome: 1) There must be a Number Czar. We cannot each run our own atalkad programs with arbitrary (usually identical) network numbering schemes in the atalkatab. This is exactly the problem with massive DECnets... HEPnet overcomes this problem by pre-assigning some range of numbers, then leaving another range for Local-Use only.... ie: the second range is some "private" range that each member institution uses as it pleases. All this overlapping number stuff really smells bad to me. Anyway, NYSERnet Inc. has offered to be the number czar for a New York State appleNet over NYSERnet bandwidth. 2) KIP and K-Star do not collate fragmented IP packets. So, if there is some zone information or other network packet (like from a document going to a remote laser printer) that gets broken up by an intermediate device, the receiving Kinetics box will not be able to reassemble the packet from the pieces. Now, there have been two messages on info-appletalk (one from Patrick Gaughan (pgaughan@nmsu.edu) and one from Phil Farrell (farrell@erebus.stanford.edu) that say (I paraphrase): If all you want to do is laser print, just use the laserwriter spooling program from the Columbia Appletalk package-- lwsrv. Basically, define a spooler on your local Unix machine that is visible to the local appletalk. However, instead of printing on a local printer, it sends the print job to a remote spooler program on a remote Unix machine. The remote program prints the file onto the remote Apple LaserWriter. This way, any IP packet reassembly is done by the intermediate Unix machines. Additionally, you don't have to wait on the initiating Mac for the potentially large job to be completed, and, you don't have to worry about laserwriter timeouts. This is pretty cool stuff, and it may be the greatest amount of functionality we can hope to achieve until there is some cure for the fragmentation problem. I am told that proteon is working on an Appletalk interface board that would fit into one of its routers. I am told that Cayman is working on correcting the same problem on their GatorBox. However, I have not gotten to the point of even seeing the problem, nevermind worrying about the fix. 3) Protection is unkown-- While we live in an open world where everyone loves each other and does not unto others as they would not want done unto themselves, there are martians and other evil extraterrestrials that would see the Chancellor's printer on a zone and print pornography on it. Of course, this possibility exists within our campus now, but we're talking accountability. Basically, I am concerned that Kinetics does not support even the meager zone protection scheme of KIP with their K-Star program. (If you are doing something about this, Kinetics, Please respond!) 4) Performance/timing problems-- I am not quite sure about this, but I am told that timeouts built-in to current appletalk programs reflect a fairly simple model. This seems to be adequate for our campus wide appletalk internet, however, even here, some slow unix machines and routers cause less than perfect behavior. I expect this to be magnified if/when we introduce the added variance in delays caused by traversing a whole bunch of other systems (internet router, shared T1 link, remote router, remote ethernet, etc.). I hope this helps. Please keep your work posted, as there is much interest in such stuff. -Samuel Asher -University of Rochester
billn@kinetics.UUCP (Bill Northlich) (01/31/89)
In article <8901251841.6201@uhura.cc.rochester.edu> asher@cc.rochester.edu writes: > Basically, I am concerned that Kinetics does not support even the > meager zone protection scheme of KIP with their K-Star program. > (If you are doing something about this, Kinetics, Please respond!) It's in beta now. Release targeted for March. Also has other desireable features. /b