On Wed, Jan 31, 2001 at 10:03:28AM -0500, Alberto Accomazzi wrote:
...
> I'm not sure why the patch never made it into the main rsync package
> (nor do I know if it was ever submitted to tridge, I certainly didn't
> see anything on the list about this)
...
I found a saved private message dated July 14, 1999, from Tridge to Bert
Dempsey that I was on the Cc in which Tridge said:
I've just read your paper and I think that it is certainly a
worthwhile approach. The only problems I see are:
- catching inconsistency between clients. what happens when (for some
outside reason) two clients get out of sync. the delta info can't be
right for both clients.
- what happens when rsync decides to be multi-pass. It is only
multi-pass _very_ rarely, but it can happen now that we use
shortened block checksums and rely on the whole file signature to
trigger a resend. That doesn't seem to fit well with rsync+.
I was very glad to see your comments on minimal impact on the rsync
code in your paper! Thanks for thinking of that and taking the extra
time that such consideration takes. Are the current rsync+ patches up
somewhere for me to take a look at? I'd like to see whether they
should go into the cvs tree.
He went on to point Bert to librsync. As far as I can tell that was the end
of the discussion because I don't have any saved message.
- Dave Dykstra
