Nathaniel Smith <njs@xxxxxxxxx> writes:
> On Mon, Mar 21, 2005 at 07:22:05PM +0900, Jun OKAJIMA wrote:
>> One idea.
>> Monotone uses hash to identify a file. Rsync uses hash for the same purpose.
>> By using rsync for checking in files and if rsync can pass hash values to
>> monotone, there is no need to recalc hashes.
>
> I'm afraid I don't quite understand what you mean by "using rsync for
> checking in files". The only thing that monotone does that is similar
> to rsync is our "netsync" protocol for transmitting changes, and
> netsync is much more efficient than rsync. So I don't know why we
> would want to use rsync for anything.
Didn't monotone once use rsync in some way? (Before I first saw
monotone, so I've no idea how. Maybe netsync started as a variant of
rsync?)
Anyway, even if monotone used rsync, and could get file hashes from
it, that would just halve the number of hashes in some cases. My
guess is that hashing files isn't a significant cost in most cases.
(i.e., there are much better places to look for optimization.)
[...]
|
