On Sun, Mar 11, 2001 at 08:52:25PM -0500, Andre John Mas wrote:
> Martin Pool wrote:
> > 
> > On 11 Mar 2001, Andre John Mas <[EMAIL PROTECTED]> wrote:
> > > I am looking at possibly doing an rsync port to Java, depending on
> > > time constraints, and I would like to know whether anyone has already
> > > set up such an effort. If there is one already under way, then it
> > > wouldn't make sense for me to duplicate the effort.
> > 
> > I don't think there is one yet.  You're welcome to do it.  I think
> > Java would work quite well for this sort of thing.  There is some
> > example code in other languages (e.g. pysync) that may guide you.
> 
> I will look at the python code, as suggested, as this should help
> me understand the necessary components. 

A little warning about the pysync... it was developed without attempting to
be compatible in any way with existing rsync/xdelta/libhsync etc. It
performs the same algorithms as them, but the data formats take advantage of
python data structures. It also favors algorithmic cleanness over
performance tweaks, for example it uses a pure adler32 for the rolling
checksum instead of the hybrids used by rsync and xdelta.

It also has zero network-transport stuff; it just calculates signatures,
calculates deltas, and applies them. Any network transport stuff is left to
the reader as a simple excersize :-)

This has it's advantages; the code is very simple. It is now just over 500
lines, of which nearly 50% are comments, and this includes both rsync and
xdelta style deltas. Actualy, the xdelta delta stuff is new... I'll put up a
new release that has it in it right now. I've also done some tweaks that
have made it 33% faster than version 1.2...

-- 
----------------------------------------------------------------------
ABO: finger [EMAIL PROTECTED] for more info, including pgp key
----------------------------------------------------------------------

Reply via email to