[comp.protocols.appletalk] AppleShare Woe Solution Found

Ravinder.Chandhok@CS.CMU.EDU (Rob Chandhok) (04/01/89)

Well, it turns out that the right solution is to get a KFPS-4, which won't
drop the packets in the first place.  This doesn't mean I think that ATP
isn't at fault here, AppleTalk should be able to deal with unreliable
networks better than that!

Here are some numbers, for those of you who like to see real data.  The file
I was transferring was MacWrite 4.6, which is 78K bytes.  The file was being
written FROM an ethertalk host, through a KBOX running KIP 0688, and onto a
server on LocalTalk.  The last time is for a Mac II on LocalTalk copying the
same file to the same server.  The server is a Mac Plus with HD20's.


Situation:          Actual Time:          Bytes/Sec:       Bytes/minute:
--------------------------------------------------------------------------
KFPS-2              270 secs (4:30 min)     288                17K
KFPS-4               22 secs               3545               212K
LocalTalk Only       13 secs               6000               360K


So, you can see that with good gateway hardware, we can achieve good results.
I still want to play around with patching the retry timer, though.
In addition, printing from out EtherTalk hosts to the LW ntx just got much
better, also at least an order of magnitude faster.

Rob

ps: can anyone tell me why KIP reports only 19 buffers on bootup?  I thought
the KFPS-4's had scads more memory than that!