On Dec 4, 2007 11:20 AM, Richard L. Hamilton <[EMAIL PROTECTED]> wrote:
> Would this be one of the tools that might eventually help getting labelling 
> out
> of device drivers and filesystems, and into a set of common functions that 
> just
> pass back the partition info to the driver (or the size of the reserved area 
> at
> the beginning of the slice to the filesystem)?
>
> I've always hated that quite a bit of disk label knowledge is
> embedded in drivers and filesystems, rather than in separate support
> routines (which could enable smarter or more consistent (between x86 and
> SPARC)) handling of trickier cases like logical partitions and so on.  To the
> extent that drivers could delegate label parsing, partition setup and
> corresponding minor device creation to support routines, they'd surely
> be simplified.  And think about what pcfs has to know about FDISK
> partitions: if someone wanted to create a native (not just FUSE) port of a
> pre-existing PC filesystem (ntfs, ext2fs, whatever), would it also have
> to be burdened with such knowledge?
>
> I can understand that access to media is essential, and has been tweaked
> quite a bit.  But if that continues to excuse aggregation rather than design,
> it will IMO all just collapse under its own weight eventually.  And I find
> it hard to believe that the present approach is anywhere near as well
> layered as it could or should be.
>

Richard,

I pretty much share your feeling about current labeling state
of affair. However, I do not see how this project can address the
problem you are describing in an elegant way.
I am, nonetheless, open to suggestions in this direction.

-- 
Regards,
        Cyril
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss

Reply via email to