I'd be surprised if "gzip --rsyncable" simply flushed the buffers
periodically. That would indeed be ineffective in the face of
insertions or deletions.
The method suggested by the author of rsync (Andrew Tridgell,
Efficient Algorithms for Sorting and Synchronization, PhD Thesis,
Australian National University, Feb 1999, page 77, Section 4.4.2) is
that a b-bit rolling signature be kept of a small window, and the
buffers be flushed when the signature hits some particular value.
This flushes every 2^b bytes on average, and is robust against
insertions and deletions. (It is probably also faster than a
Manber-hash, and could be piggy-backed onto the existing CRC code.)
The phrasing of the man page,
--rsyncable
While compressing, synchronize the output occasionally
based on the input.
is consistent with this, as it says "occasionally" rather than
"regularly" or "periodically" or "every 4k", and moreover flushing
every 4k would not be "based on the input."
Of course, it would be easy enough to test ...
--
Barak A. Pearlmutter <[EMAIL PROTECTED]>
Hamilton Institute & Dept Comp Sci, NUI Maynooth, Co. Kildare, Ireland
http://www-bcl.cs.nuim.ie/~barak/
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
