RE: GPT byte-swapping patches

[Matt_Domsch]

> > Due to Intel madness, RFC-defined UUIDs and GUIDs and 
> > Intel's specification of efi_guid_t aren't quite the same.
> 
> Have you tried attacking them with a cluebat?

Yes.  I've been staunchly ignored.  Big-endian?  What's that? :-)
Even if Intel admits they screwed up, they're unlikely to change their
implementation to match their specification...
 
> Anyway, UUIDs are generally used in a rather opaque manner, 
> so it doesn't really matter that much?

It shouldn't, and it doesn't (much), which is why it hasn't been a problem
for the past year and a half since I started all this GPT nonsense.  But,
for correctness, since I'm using uuid_generate() to make a uuid, but the
efi_guid_t format has those fields byte-swapped, it's better to fix it now.
The only real issue is the miniscule possibility that two uuids, one in uuid
format, and one in efi_guid_t format, wind up being the same value when they
were "guaranteed" to be unique.

We're trying to figure out the right way to fix this in the Linux kernel
proper too.  There's a desire on one hand to treat everything as uuids, and
make efi_guid_t go away entirely.  But, then you've got to keep knowledge of
state in your algorithms to maintain correctness - I'm leaning toward
leaving efi_guid_t in its broken form.

-Matt

--
Matt Domsch
Sr. Software Engineer
Dell Linux Solutions www.dell.com/linux
#1 US Linux Server provider with 24.5% (IDC Dec 2001)
#2 Worldwide Linux Server provider with 18.2% (IDC Dec 2001)