CompuServe Thread

#File protocols

4 messages in this thread
#8699From: Larry Phillips/ICUGFeb 18, 1986 3:14 AM
The ONLY thing Kermit has going for it is the fact that it is implemented on a large number of machines. Personally I prefer spending 10 minutes downloading and 1 minute CHOPing to spending 20 minutes downloading. THE best protocol may not have been implemented yet, as all I have seen have a few snags to catch you. One thing is certain though, any protocol that doesn't approach the maximum theoretical transfer rate to within a few percent is NOT a contender for being called evn good, let alone best. Regards, Larry. * Replies: 8711, 8723
#8711From: Chris BaileyFeb 18, 1986 8:44 AM
Here's a suggestion for a better file transfer protecol. It came to mind after reading all the messages concerning CHOP confusion. Why not have a protecol that automatically squeezes a file, sends it through the phone lines, then unsqueezes it at the receiving end. Better still, send a header that contains the file name and other info that allows the receiver to specify whether the file is to be unsqueezed. This would allow the squeezed version to remain on CIS for example, then when a person downloaded the file the protecol would detect that the file was already squeezed and would not attempt to squeeze it again. In the error detection area, I also have a suggestion. In my days working on SCADA (Supervisory Control and Data Acquisition) systems we used a modified CRC that was 6 bits long for a 4 byte transmission. This CRC was flipped (ie bit 0=bit 5, 1=4, etc) so that the error detection bits were still the last 6 in the message, but were a mirror image. Apparently (I am no math whiz) this decreased the odds of an error significantly over a regular CRC. – Chris * Reply: 8712
#8712From: Stephen R. WaltonFeb 18, 1986 9:19 AM
I know that there are Kermits available for both Prime and Univacs. As for line delays, compression, and better window checking, I should have stated originally that TheSource uses a new superset of the Kermit protocol called "sliding windows", which involves sending a large number of packets (as many as 31) before waiting for an ACK. Since the ACK in Kermit is also a sequence-numbered packet, the sender may receive an ACK with sequence number 25, which is an implicit ACK of packets 1-24 as well; the next bundle of packets will include a retransmit of numbers 26 through 31, plus enough new packets to make up the next group. The new protocol also includes compression of data by the sender and uncompression by the server. Tests show 30 to 45% reductions in transmission time over either XMODEM (128 byte packets) or old Kermit (90 byte packets). –Steve
#8723From: William Volk(Aegis DevelFeb 18, 1986 12:07 PM
If AmigaBinary/Xmodem CRC has a count in the header …. you won't even have to chop … and it will be compatable with compuserve.