Re: Proposal: Checksum header or method

    Simple file length may not suffice, so it would be better to determine
    the checksum.  Although the client could calculate this, it makes more
    sense to do so on the server side, to preserve bandwidth.

The use of a simple checksum might be too prone to false positives.
I'm not an expert on the topic, but apparently people have developed
"fingerprinting" algorithms that have very low probabilities of
mapping two texts to the same value.

I found one citation on the topic:
	Rabin, M. O.
	"Fingerprinting by Random Polynomials"
	Center for Research in Computing Technology
	Harvard University
	Report TR-15-81, 1981

but I suspect there are others.

-Jeff

Received on Tuesday, 25 July 1995 13:14:18 UTC