Hi,

On Fri, Sep 15, 2000 at 12:07:32PM +0200, Ragnar Kj?rstad wrote:

> > I have noticed that several of you have mentioned wanting to have a
> > VFS-based solution.  I suppose the reason is uniformity of behavior
> > across filesystem types?  Having been through this pain already once,
> > I think you are taking on a very difficult task, for a number of reasons.
> 
> Yes, uniformity of behavior and avoiding duplicating code.

Right, as long as the common API can be made both (a) generic enough
that it can express all of the different implementations which already
exist, and at the same time (b) simple enough that the core API will
work no matter which implementation is underneath.

This is not an easy compromise.  I've CC:ed this to the linux-fsdevel
list, as there are a lot of different filesystems out there which may
be impacted by this and not everybody involved is necessarily on the
DMAPI list.  

btw, there was a _long_ thread about this a while ago on l-k, where
Linus expressed an opinion that treating files as directories was a
good initial cut at an extended attribute API --- ie, allow
"file/named-attribute-stream" to be opened even if "file" is a regular
file.  The only snag with this is attributes on directories, which you
still need for things like ACLs, but you can fix that with an "fdopen"
which names a file relative to a given fd, rather than by a pathname;
so you can open a directory and fdopen() the attributes if you want
to.

Cheers,
 Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]

Reply via email to