Marc Perkel wrote:
Nikita Danilov wrote:
Well, this song has been sung many times already. :-) Hans has strong
opinion about xattr API (which, indeed, is not a paragon of design
beauty by a large margin).
I'm new to this debate. What is his strong opinion?
Well, I'll let him speak for himself, but I think it's basically this:
xattrs suck, for reasons which are varied and irrelevant.
When it's done, the sys_reiser4 api (and file-as-dir/metas/whatever)
will be able to do everything xattrs did, only much better, faster, and
more powerfully.
So, it's more worth their time to work on other things, like:
- kernel acceptance
- repacker
- crypto/compression
- security ("views" or ACLs)
Of course, if someone wanted to pay them for xattrs, and they didn't
have another contract going, they'd probably support xattrs. But for
now, it's probably up to us to add xattr support.
Once it's added, these should do the same thing:
openat ("some_attribute")
open ("metas/some_attribute")
// or maybe
open ("metas/xattrs/some_attribute")
Well, there's more to it than that, but the most important thing is, the
xattrs should be a subset of the metas namespace.
