Transferring large files using rsync

jw schultz jw at
Fri Sep 12 16:48:37 EST 2003

On Thu, Sep 11, 2003 at 11:04:26PM -0700, cbarratt at wrote:
> jw schultz writes:
> > On Thu, Sep 11, 2003 at 05:43:56PM -0500, Seann Herdejurgen wrote:
> > [snip]
> > > Is there a way to tell rsync to update files "in place" and not
> > > create a temporary file?
> >
> > No.  Nor can rsync be modified to do so.
> Is that true?  I thought one could readily add an --inplace option that
> would restrict the block matching algorithm to only match blocks at the
> same or later offset in the file.

Depending on block size and buffering that would be exactly
the same, or same + 1 block and later.  The only point where
in-place doesn't get complicated is when the entire file
fits in memory.  Any failure on the part of the sender to
anticipate a matched block relocation failure would be
irremediable with the current pipelined protocol.

The returns on in-place would rapidly diminish precisely on
those files where it would seem to have the greatest value.
Tar files, database tablespaces and filesystem images all
have high match rates and high rates of semi-random block

	J.W. Schultz            Pegasystems Technologies
	email address:		jw at

		Remember Cernan and Schmitt

More information about the rsync mailing list