time@oxtrap.oxtrap.UUCP (Tim Endres) (12/08/89)
I have been working on an upload facility running on a Unix system receiving from Macs. I have Telebit T2500's on both ends. After much work getting in *at* 19200 and getting flow control out of the way, I was able to get >1200 cps from the T2500 running XMODEM spoof with 1K blocks. I can get >1100 cps with standard 128 byte blocks. I am running the public domain code "rz" as "rx" on Unix. PROBLEM: I notice that the Unix side is reading *ever* so slightly slower than the Macintosh is writing. But due to the spoofing, the Mac gradually gains on Unix. Thus, with very large files (> 500K) the amount of time the Mac waits at the end for the Unix acknowledge of the "transfer completion" seems to time out and the transfer "appears" bad to the Mac. Yet, it appears to actually complete the transfer, if the modem is not hung up before Unix can complete. Small files are transferred with absolutely no problems. I notice that file transfers from one Mac to another do not have the same problem, they seem to keep up. QUESTIONS: Is the "rz" code capable of handling data rates of 19200 on a SysVr2 SLOW file system? Could the CRC checksum be too much overhead? (The Unix machine is a 68030 @16MHz so I don't think this is the bottleneck) Is the "rz" code efficient at writing to the file system? Is there anything else I should suspect (like the Unix driver)? SUMMARY: Does anyone out there have extensive experience with large file transfers over asynch modems at 19200?