2002-07-29-12:11:54 Biju Perumal: > [ why rsync's design rather than accumulating diffs ]
The approach you describe is quite often used. Look in the neighborhood of source code control systems used in distributed ways, and you'll see it all the time. I've also used it in the past for software distribution and content management. Rsync has a couple of advantages, that make it a valuable tool as it stands, and that allow it to sometimes (but not always) encroach on the turf of such version management systems. First off, if you want to send pre-generated diffs, you really have to know what version the receiver has. In general keeping track of this can be difficult, and it makes it harder to be robust in the face of updates that fail in the middle of a multi-file update process; you end up needing to try and implement your patch process in a transactional fashion, so you can back out all applied patches if you bomb in the middle. And second, they impose on the sender the requirement to keep previous versions for differencing against. Sometimes this is an awkward requirement --- think using rsync for online backups. Rsync is an efficient distributed file synchronization tool, that uses an efficient and clever distributed differencing algorithm. It's not the best tool for all jobs, but there are many jobs that it does very well. -Bennett
msg04713/pgp00000.pgp
Description: PGP signature
