superlifter design notes and a new proposal

Wayne Davison wayned at
Mon Aug 5 21:29:02 EST 2002

On Tue, 6 Aug 2002, Martin Pool wrote:
> But my main argument against it is not so much that crunched
> headers are not more byte-efficient, but rather that they are a
> premature optimization.

For superlifter, yes I agree that such a scheme is premature
optimization (since we're still in the design phase).  My rambling
about why I did what I did in rZync and quibbling over what might be
optimal in the long run apparently came across as a desire to hash
out these issues now, which is not the case.  I think we can easily
choose a simple byte protocol now and finalize (and optimize) it
later.  That is, assuming that we've decided to go with the design
that "everything that crosses the wire is a part of a message"?  I
think that such a design makes the transport problem very nicely
self-contained and easier to get right.


More information about the rsync mailing list