Re: rsync delta between binary files

Hendrik Mangels wrote:
> hi,
> 
> easy question: I need to know which size the delta is, rsync produces
> when syncing two particular (binary) files. So to speak: what will go
> over the wire, in the end. Is this what rdiff does?

This is the librsync list, not the rsync list. librsync is a library 
implementation of the rsync algorithm, and is not used by the current 
rsync application.

rdiff is a small utility that uses the librsync library to generate the 
signature of a file, caculate a delta between a signature and another 
file, and apply the delta to the file. It does not have any network 
transport, so  doesn't send anything "over the wire".

You can use rdiff to syncronise something "over the wire" using rdiff by 
using something else to transmit/receive the signature and patch files. 
This would be done as follows;

1) generate a signature for the old file to be updated
2) send the signature to the other end
3) use the signature to calculate a delta from the old file to the new file.
4) send the delta to the other end
5) apply the delta to the old file, updating it to match the new file.

Note that the signature must be sent to the updated end, and the delta 
is sent to the old end. I think the question you are asking is 
effectively "how big is the signature and delta".

The size of the signature is about 12 bytes per block. The default block 
size is 2048 bytes, so the signature size should be about 12/2048th the 
file size (Note: for very large files I suggest specifying a block size 
of sqrt(filesize)).

The size of the delta depends on how different the files are. The delta 
format includes "copy" and "insert" instructions. The copy instructions 
are very compact. The insert instructions include all the data to 
insert. I'm not sure, but rdiff may also compress this delta using gzip. 
This means that the delta pretty much only includes changes (maybe 
compressed). No changes -> small delta (one copy instruction), all 
changes -> large delta (one big insert instruction).

-- 
Donovan Baarda


-------------------------------------------------------
SF.Net email is sponsored by: GoToMeeting - the easiest way to collaborate
online with coworkers and clients while avoiding the high cost of travel and
communications. There is no equipment to buy and you can meet as often as
you want. Try it free.http://ads.osdn.com/?ad_id=7402&alloc_id=16135&op=click