Just some random ideas on this topic...
How about using more than one checksum per file?
I mean, you may say that each block of on megabyte of a file has its own
checksum
when you start reading file content to compute MD5 you may stop as soon as you
found one difference.
one other nice thing with this trick is that it may be used to transfer only the
changed blocks of a file...
this is a matter of concern for big binary files...
maybe take a look at how rsync works?
(I must admit I don't know if cvs uses clever algorithms to transfert only
changed parts of files during commits/updates on low BW links or big files)
Olivier
vvvvvvvvvvvvvvvv sorry for this garbage vvvvvvvvvvvvvvvvvvv
