Re: [Duplicity-talk] zfec vs. par2 (and, hello there!)

[zooko]

On Nov 4, 2008, at 4:34 AM, address@hidden wrote:

> At least the command line interface do not support to file the  
> redundancy information separately... it always creates a set of  
> files, of which a definable count is necessary to recreate the  
> source file.

In that set of files, the first 3 of them actually contain the  
contents of the input file (plus each file has a couple of bytes of  
header).  You can inspect the files that zfec produces to see what I  
mean.  Make a file with contents "0123456789" and then run "zfec" on  
it.  Then run "less" on the resulting output files.

> only one file can be processed at a time.

It would be easy to change the cmdline zfec tool to process multiple  
files, but what should it do with them?  I guess maybe "for FILE in a  
b c d e ; do zfec $FILE ; done" or else "tar cjf a.tar.bz2 a b c d e  
&& zfec a.tar.bz2" would probably be the best way to handle multiple  
files.

> I don't know about the inner API as I am not python literate ...  
> regards ede

The zfec README.txt describes the API.  There is a function called  
"encode()".  You tell it which block numbers to produce.  If you tell  
it numbers which are >= K then it produces only "check blocks" a.k.a.  
"parity blocks" a.k.a. "secondary blocks", and produces no file data  
blocks.

http://allmydata.org/trac/zfec/browser/zfec/README.txt

Regards,

Zooko
---
http://allmydata.org -- Tahoe, the Least-Authority Filesystem
http://allmydata.com -- back up all your files for $10/month