Syslinux - Apr 2016 - Confusion regarding cpu-arch values for EFI ByteCode and 64-bit

This document:
http://www.iana.org/assignments/dhcpv6-parameters/dhcpv6-parameters.xml

Has the same (plus expanded) list of cpu arch values for parameter 0x93.
This list matches RFC 4578 except for values 7 and 9 which are EFI ByteCode
and x86-64.

Turns out RFC 4578 is wrong

https://www.rfc-editor.org/errata_search.php?rfc=4578

And 00:07 *is* the correct value for 64-bit x86 clients and the comment on
the wiki should be updated.  See:

http://www.syslinux.org/wiki/index.php?title=PXELINUX

That said, because RFC 4578 is still shown incorrectly unless you click on
the "ERRATA" we will probably see clients written with the
"wrong" values
for quite some time and we should probably continue to recommend that both
7 and 9 be accepted as efi 64-bit and assume that no one would ever
knowingly use ByteCode...

Just FYI,
Phil P.


-- 
Philip Pokorny, RHCE
Chief Technology Officer
PENGUIN COMPUTING, Inc
www.penguincomputing.com

Changing the world through technical innovation

> This document:
> http://www.iana.org/assignments/dhcpv6-parameters/dhcpv6-parameters.xml
> 
> Has the same (plus expanded) list of cpu arch values for parameter 0x93.
> This list matches RFC 4578 except for values 7 and 9 which are EFI ByteCode
> and x86-64.
> 
> Turns out RFC 4578 is wrong
> 
> https://www.rfc-editor.org/errata_search.php?rfc=4578
> 
> And 00:07 *is* the correct value for 64-bit x86 clients and the comment on
> the wiki should be updated.  See:
> 
> http://www.syslinux.org/wiki/index.php?title=PXELINUX
> 
> That said, because RFC 4578 is still shown incorrectly unless you click on
> the "ERRATA" we will probably see clients written with the
"wrong" values
> for quite some time and we should probably continue to recommend that both
> 7 and 9 be accepted as efi 64-bit and assume that no one would ever
> knowingly use ByteCode...
> 
> Just FYI,
> Phil P.
 
So, someone "decided" (during 2016Feb) that RFC 4578 is the
"wrong"
one, instead of deciding that the other documents (that were composed 
at a later date than RFC 4578) were "wrong"? Just like that? Isn't
this
a "because I say so"?

What if someone else would had decided to propose an errata to the 
_other_ documents, preserving the original RFC 4578 list?

The note in the PXELINUX wiki page about the misused values was 
originated from a document / presentation:
IETF 72, "IPv6 Remote Boot Requirements From UEFI Forum", by Dave 
Thaler & Ken Grewal on behalf of Vincent Zimmer (from Intel).

So, someone could validly claim that the errata should be introduced 
for the other RFCs and the IANA document, and that some implementations 
(and some documents) are wrong, but not the original RFC 4578.

I'm not sure even the UEFI Forum would be the adequate entity to 
coherently decide which values are the "correct" ones.

In practice, I would agree that all the DHCP conditionals will be 
needed for a long time, and that finding some practical conflicting 
cases will not be a surprise. Therefore, the note in the wiki page 
should be left "as-is" for now.

FWIW, I raised the matter in this Syslinux Mailing List during October 
2014. See "DHCP option 93 for UEFI" (and the related emails following 
it):
 http://www.syslinux.org/archives/2014-October/022682.html 

Regards,
Ady.