extended attribute standardization

[Jos van den Oever]

2006/11/27, Pete Ryland <pdr at pdr.cx>:
> I've been thinking of good uses of xattrs and discussing this
> elsewhere, and we came up with a few good ideas.  One is to cache data
> that are otherwise very expensive to generate, like md5sums and rsync
> block checksums.  The only drawback is making sure these are kept in
> sync, especially if the md5sum is used as a security measure.  This
> would require these xattrs to be automatically removed by the
> filesystem in the case of the file changing, perhaps by being put in a
> separate namespace.

In Strigi, one can calculate the sha1 for each file. This will be
stored in the index and enable quick searching for duplicate files.
When the md5sum / sha1 is stored in an extended attribute, quick
searching is not possible unless the value as indexed in addition.
This takes more diskspace and also adds a second synchronization
problem. In Strigi the synchronization is checked by storing the mtime
of the file along with the filename and the checksum.

Cheers,
Jos