Re: Announcing xorriso-0.3.6
Hi,
Bill Davidsen wrote:
> As long as you can back up and restore what is there in Linux now,
> and use features which other applications and OS are required to
> ignore if they do not understand, then you have provided a useful
> backup capability which is "portable enough" to be useful.
That is quite exactly my own justification
of AAIP: it adds value and does no harm.
(Except the blooper with the AA field which
is to be repaired soon.)
> Rarely are
> the xattr needed to be moved to other OS, because the exact
> functionality of the xattr may not be precisely the same.
Yes.
For now xorriso records only xattr from
"user." namespace but not from "system.".
libisofs could, but xorriso says it shall not.
It might be dangerous to carry them from
one disk to another. It also demands
superuser authority to set those attributes
to disk files. (Superuser is dangerous in
itself.)
Maybe later, as extra option, if somebody
finds a good use case for that.
Therefore it was necessary to define an
own ACL format because this concept is quite
portable in the X/Open world.
AAIP ACL are about half of the size of
the according Linux-ReiserFS "system." attribute.
Compact storage representation is important
with AAIP because large attributes need
extra data blocks which cause extra random
access moves of the CD/DVD/BD read head.
Another appeal of AAIP is that the ISO producer
software can store information for internal
purposes in the "isofs." namespace.
I already defined an attribute for the original
dev_t and ino_t of files: "isofs.di". This helps
with incremental backups.
"isofs.cs" records the character set which was
used when producing the name tree of the image.
Similar attributes will be of help when xorriso
will learn to encrypt and compress file contents.
Have a nice day :)
Thomas
Reply to: