[ovs-dev] [PATCH 15/17] nx-match: Add support for experimenter OXM.

[Ben Pfaff]

On Wed, Sep 24, 2014 at 10:57:08AM +0900, YAMAMOTO Takashi wrote:
> > Testing experimenter OXM is tricky because I do not know of any in
> > widespread use.  Two ONF proposals use experimenter OXMs: EXT-256 and
> > EXT-233.  EXT-256 is not suitable to implement for testing because its use
> > of experimenter OXM is wrong and will be changed.  EXT-233 is not suitable
> 
> sorry for sidetracking, how EXT-256 is wrong and will be changed?
> i'm interested because i implemented the extension in Ryu.

The original design for OXM had 32-bit standard OXMs and 64-bit
experimenter OXMs.  Then, later on, somehow this design was forgotten,
and experimenter OXMs were re-envisioned as 72, 80, 88, or 96 bits
long, where the additional 8 to 32 bits provided additional
experimenter-defined field information.  This is difficult to handle
because the length of the additional bits is experimenter-specific, so
that a switch or controller cannot figure out whether two fields are
even the same unless it knows how long the additional bits are for
some specific experimenter.  It is also not explained in the standard
anywhere.  I don't know where it came from, but it seems like a
mistake and will be rectified by reissuing a corrected EXT-256.  It
should not be necessary to actually change the OpenFlow standard
otherwise, since the standard didn't actually say to do anything this
way.