I have been following this thread.
I am working on rsync for an embedded application, but it has nothing to do with
program loading.
Donovan recently provided some formulas on figuring out the required checksum
size relative to file size and acceptable failure rate.
In the formulas, he assumes that the block size is square root of file size.
I have done some benchmarking of rsync and for large files (ie. 1 -55 Gigbytes
for example), the block size is larger than a square root of file size. This is
based upon the xxxx.rsync_csums file size produced when you --write-batch=xxxx.
I have done some calculations assuming that a check sum per block is 20 bytes
(32 bits + 128 bits) and 24 bytes (assuming there is an extra 32 bits of
overhead data) and the number of checksums appears to be square root of file
size divided by a number between 2 and 3.
This observation isnt central to what I am doing, so I havent really tried to
pin this down.
I thought that I should bring it to the attention of anyone following the
thread.
wally
