I've released the next version of my rZync test app. You can find a
link to it here:
http://www.clari.net/~wayne/new-protocol.html
You should also snag the referenced librsync source, as some important
bugfixes in librsync are needed to compile rzync.
For those that don't know, rZync is my new-protocol test app that I'm
using to try out some ideas on how to improve the rsync protocol. It
transfers directory information incrementally, so it should have a much
lower memory overhead than rsync.
The most important change in this release is that I've replaced the
name-cache code with something that will be more robust and should work
great with really large file transfers. I also changed the command-
line syntax and have it now parse several new options, such as -r, -p,
-t and such (i.e. the previous behavior of -a being hard-wired to "on"
is no longer present) and a few other things. Another important bug fix
closes a neglected file handle so we don't overflow the open file limit.
** Be sure to use the new "rs" controlling script and not the old one.
**
I've tried the code out on a fairly large data set (~4000 files in ~500
directories), but nothing close to some of you million-file folks. I
would not yet recommend trying rZync in a production environment, but if
you can run some large file-count tests, please let me know how things
go.
..wayne..
