Tom Jennings
     Fido #1
     9 May 85

             This is a copy of a letter I sent to Inforworld on
     an article that appeared. I'm submitting it to the
     newsletter since it seems that Infoworld won't publish it.
     (Or, maybe I just didnt wait long enough.)


     InfoWorld
     Attention: Kevin Strehlo
     Re: "Modem Battle Brewing" article, 29 April


     Tom Jennings
     2269 Market St. #118
     San Francisco CA 94114

             Here's some stuff on that article, I just had to
     respond. Thanks for reading so far. If you have any
     questions, comments or whatever you can call me at the above
     number any day from 8 AM til midnight.


             There were some serious errors made in the article
     "MODEM BATTLE BREWING" on page 16 of the 29 April issue on
     the XMODEM protocol.

             Before talking about the specific problems, a
     misconception needs to be cleared up. The word "XMODEM"
     refers to both the protocol and is also the name of a
     specific CP/M program for remote bulletin boards. "XMODEM"
     is now taken to mean the binary protocol, not the name of a
     specific program residing on someones disk. This can be very
     confusing, but unless you are talking about a specific copy
     of a specific program, Xmodem refers to the protocol.

             It is completely untrue that Xmodem cannot run
     beyond 1200 baud, and equally untrue that it does not detect
     errors reliably.

             The original Xmodem program, XMODEM, MODEM7 and all
     it's variants, have a software bug that prevents them from
     operating at above 300 baud. It is a bug in the
     implementation, not the protocol, and has been there for
     years. My implementations of Xmodem (and many others') have
     been used for over three years at up to 38,400 baud (as fast
     as my serial ports go) with no problems whatsoever. The same
     program can communicate to any original XMODEM program on
     MSDOS or CP/M. In the FidoNet Bulletin Board Network, a
     public domain packet switching system consisting of more
     than 250 computers in the US and Europe, many nodes are
     using 2400 baud modems quite reliably, thank you.

             The error checking isn't quite as bad as many would
     lead you to believe. The original program by Ward
     Christensen had only a simple checksum error check; later
     versions implement a CCITT CRC-16 error check. Also, there
     is some doubt as to whether CRC error checking is all that
     much better than checksums over dialup telephone lines;
     while it is true that adjacent single bit errors can get
     through Xmodem's checksum process, errors on phone lines are
     almost always burst errors (that is, many bytes are trashed
     by noise on the telephone), these are very rare, and so in
     "real life" Xmodem is quite reliable. In any case, CRC-16 is
     available. In any case, the worst case error rate is
     something like 0.035% using checksums. No protocol can
     guarentee 100% accuracy, no matter what they may claim.

             Xmodem has many faults, such as half duplex
     operation and requiring an eight-bit data channel (can't be
     used through Telenet or TymNet...), but it extremely easy to
     implement, and is very reliable if done right. It is surely
     far from perfect, but is very suitable for small computer
     use where the constraints don't matter.

             You will also notice that the biggest detractors are
     firms with commercial interest in their own protocols. As to
     the importance of the Xmodem protocol in the real world,
     keep in mind that the modem manufacturers (and many others)
     who came out with software products with "improved" (read:
     incompatible) protocols were forced by the micro marketplace
     to retrofit Xmodem. Two examples are SMARTCOM and CROSSTALK.

             Ward Christensens protocol has problems, as he will
     admit, but it is still very strong and usable in the current
     environment. There are probably 10,000 bulletin board
     systems in the US that support Xmodem, and probably ten
     times or more individuals that use Xmodem with bulletin
     boards or for point to point use. It may be an "underground"
     protocol, but it works well in practice, and introducing new
     protocols will have to take this huge base of users into
     account.


     Ref: FidoNews 2-15 (27-May-1985)
     http://195.226.109.55/jhassler/wif/doks/fnews/fido215.txt
     By: Tom Jennings