KTar patch to fix speed problems with tar ioslave

[Jan Schäfer]

Hi Alex,

> Well, without the temp file it would be nicer. I didn't look at the
> implementation of the tar/gz/bz2 ioslave.
> What's is actually taking so much time ? (I know, the decompression).
> ioslaves have a "special()" member, which you can use to implement
> non-standard things. Would it be possible to tune the tar and/or the gz/bz2
> so that they have to seek through the whole file at first and then cache
> the offsets where the file starts so that for the second file they know
> already where the data starts which has to be uncompressed ? Is this
> possible ?
No, I do not think that it is possible to tune the compression filters.
The compression filters are based on external libraries and I think that 
compression formats in general are not designed to have an efficient random 
access to the decompressed data, but to decompress the whole data at once.
Also every compression filter had to be improved. My solution is independent 
of the compression filters and does not need much code changes.

Jan