Date: Wed, 19 Mar 86 17:00:47 cet To: INFO-KERMIT@CU20B.ARPA From: FI%NORUNIT.BITNET@WISCVM.WISC.EDU Subject: Sperry Kermit We are running the Sperry Kermit by Paul Stevens, dated june/july 1985. If anyone is interested, here is a report of some of our problems with this Kermit, and our fixes for them. Apart from these minor annoyances, the MASM Kermit has been a pleasure to use! (1) There is no need for Kermit to assign the Sperry work file exclusively, apart from the risk that someone else writes to the file during transmission. To me, this was more annoying than useful, so I changed the file assignment as shown. (2) As distributed, Kermit will not treat program file elements with multiple cycles (indicated by fieldata S in S3 of the label control word), unless the data part of the label conforms to the SDF standard (*SDFF* in first word). As a result, elements written by the system line editor ED, will not be transmitted correctly. That is, if our fix is not applied... (3) When ACK'ing a previous data packet, Kermit as distributed put the first 6 characters of the ACK'ed packet into the data part of the ACK. I haven't seen any Kermit do that before, but it looks straight enough. However, after receiving a couple of those 'long ACKs', IBM PC-Kermit (2.28) fills the next one or two packets with garbage (typically, a lot of zeros - nicely encoded, though, so the receptor does not notice). The result is an apparently successful transmission, with a few 'black holes' in the element on the Sperry host. Changing the data size to zero in these ACKs seemed to eliminate the problem. These are the fixes in Sperry correction card format (1) -3177,3177 sTrng '@ASG,A K$E$R$M$I$T$ . ' . Not exclusive (2) -4287,4288 (3) -4713,4715 sz,h2 prline . Do a normal ack of length 0 l,u a2,prline . Seems to confuse PC-Kermit -fi Frithjov Iversen Trondheim University Computing Center, Norway ------------------------------ From: ROGER @ UK.AC.TPB Subject: Bugs in Sperry Univac Kermit Keywords: Sperry Kermit, Univac Kermit I recently acquired a copy of Sperry UNIVAC KERMIT written in assembler, for use on a non front end site. After a little tinkering , to get it to work with our setup , I discovered a couple of little coding bugs. I must admit I'm not the world's greatest programmer in @MASM, but I THINK (underline that in italics) I've sorted them out: There was a bug in the SHOW SEND routine that caused the SEND STARTOFPACKET displayed to be the RECEIVE STARTOFPACKET, and a bug in the SHOW RECEIVE routine that caused it to display the SEND STARTOFPACKET. No prizes for guessing what had happened !!!! The actual parameters in the SHOW list had been juxtaposed; simply swapping over lines 2281 and 2300 in the original source should cure the problem. Another more serious problem was that when assembled with MDLFE=0 and DCPFE=1, the code still expected to find a couple of entrypoints that weren't there: they'd not been assembled because of a conditional directive. My cure is rather elegant, but as I've no idea what I've done it may not be the right one. All I did was to move the offending reference, in line 2613 to only occur in the conditional directive immediately following it. that is line 2613 was inserted AFTER the IF MDLFE statement. That seems to have cured it, it now @MASM's without errors, and @MAP's without errors. I've succesfully used it in SERVER mode with a PC clone running CROSSTALK, so I assume I've done the right thing. If anyone else has any tips or points Id like to hear from them. Jason LoCascio, British Gas PLC 59 Bryanston Street LONDON W1 (01) 723-7030 ext. 1289 Or I can be contacted at THAMES POLYTECHNIC , via JANET :- ROGER @ 000045399000.TPB.SPCP.FTP.MAIL (We are not registered in NRS yet) ------------------------------