ari@riacs.EDU (Ari Ollikainen) (12/24/87)
I believe that Vitalink has actually applied for a patent on what they call End-to-End FCS. Their first generation of TransLAN product did NOT carry the Ethernet CRC across the serial link. Back in 1985 when we made our earliest experiments using Vitalink bridges across a point-to-point satellite link we had some DECnet users complain about garbaged files. It turned out that the DECnet hosts assumed that, since they were operating on an ethernet, the Ether CRC was sufficient protection against corruption. And we, using FTP/TCP/IP, didn't have any realy problem EXCEPT that the transmission times for files seemed to vary more than could be explained by the load on the communicating hosts. It turned out that we were using some RF modems which were 1) sensetive to RFI/EMI, and 2) used the same polynomial for "scrambling" that was used to compute the CRC on the HDLC frame that Vitalink used on the p-t-p link. Some frames were sufficiently corrupted to pass the CRC re-computation at the receiving end, thus these frames would have a NEW Ether CRC generated for them and be sent forth to fill the DEC users' file. OF COURSE, the TCP (and IP header cksums) caught the garbage storms and prevented the Internet protocols suite users from suffering file corruption!