Re: Listing the deltas

[Colin Percival <cperciva@tarsnap.com>]

On 06/06/12 21:21, Ted Unangst wrote:
> On Wed, Jun 06, 2012 at 20:24, Colin Percival wrote:
>> On 06/06/12 05:19, Tom Limoncelli wrote:
>>> On a machine where very little changes each day, there always seems to
>>> be 20 MB of "new data".  Is there a way to list which files changed
>>> and were backuped?
>>
>> No.  The way the tarsnap code works internally, it generates a complete
>> archive and then figures out which blocks duplicate data previously
>> uploaded; the deduplication code doesn't have any way of knowing which
>> files data originally belonged to (and in fact there will be blocks
>> with data aggregated from multiple files).
> 
> If I remember correctly, tarsnap is only looking at aligned blocks,
> too.

You do not remember correctly.  Tarsnap uses blocks varying between
16kB and 256 kB (average size 64 kB), with no particular alignment.

> And is the block size 8k?  (As I recall, prepending 4k of new
> data to an existing 8k file will result in 12k of new data, since the
> already uploaded data does not match the new file's block alignment.)

For that case, yes.  But prepending 4k to a 1 MB file will usually only
result in one new block.

-- 
Colin Percival
Security Officer Emeritus, FreeBSD | The power to serve
Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid