tor 2007-06-07 klockan 18.12 skrev Jason Thorpe:
> 
> On Jun 7, 2007, at 8:41 AM, Johnny Billquist wrote:
> 
> > I take it that this means you intend to switch them over to use  
> > GPTs. So
> > we'll "create" a format there, which I guess was designed for another
> > type of systems...
> 
> We don't be "creating" a format anywhere.  We'll be using a de facto  
> standard format - GPT.  GPT was not designed for any type of system in  
> particular.  It's completely general.
> 
> Using your logic, we "created" a format on x86 PCs by using BSD  
> disklabels there (their original usage was on VAX and Tahoe systems,  
> and some of the fields in the label reflect that heritage ... does  
> that mean BSD disklabels were "designed for" those types of systems  
> and thus shouldn't be used elsewhere?)

Certainly we got a bunch of headaches when we adopted the BSD disklabels
for the x86 PC, which have their MBR table. I don't think even you can
deny that this caused problems, headaches and confusion.

It might have been better to not use disklabels there, if we wanted
MBRs. You can run the system on a PC without an MBR, in which case it
makes more sense to have a BSD disklabel.

> > Oh well, as long as it keeps as non-intrusive as
> > possible. BSD disklabels are rather nice, since they are rather  
> > simple,
> > and the design makes a lot of sense, when you don't have to mix in the
> > IBM PC MBR stuff, or other similar schemes.
> > The design on a PC, with the d-partition has never felt good to  
> > me. :-)
> > But I'm onld enough to remember the time when the partition layout was
> > hardcoded in the driver.
> 
> You're falling into the trap of confusing the on-disk representation  
> with the in-memory representation.  This is what wedges is designed to  
> get around.

Huh? The d-partition on an x86 PC is visible everywhere.
And hardcoded disklabels were certainly just an in-memory
representation. There wasn't any on-disk representation back then.
That's why the disklabels came about. To move the in-memory
representation away from being hardcoded, and instead become a
reflection of the on-disk representation.

> Just so everyone is clear -- wedges is what NetBSD is adopting (albeit  
> slowly) for the in-memory representation and user presentation of disk  
> partitions.  They will be used for all on-disk partition table formats.

So if they will be used for all on-disk partition table formats, and
they are used for the in-memory representation, in which way are they
separate from each other?
(I think I know what the idea of wedges are, but to be honest I'm just
guessing. I've not bothered with following this thread, or the general
development. Apart from the fact that I'm running ancient hardware, and
mostly bother with issues that noone else really cares about, I also
have the impression that people aren't really that interested in the
views of others, such as me in general either.)

> And, FWIW, GPT partition tables are really no more complicated than  
> BSD disklabel partition tables.  the

Did you send this by accident before you finished the sentence? :-)

	Johnny