> 
 > >  > For pragmatic reasons the precise interaction between using this
 > >  > option and the behavior of automatic source address selection is
 > >  > explicitly not part of the Stable behavior defined by this case.  In
 > >  > short: automatic source address selection is not needed by any of the
 > >  > initial users of this option, it's not immediately obvious what
 > >  > "correct" is, and several open RFE's against zones may require changes
 > >  > or clarifications to this interaction.
 > > 
 > > As per the Stable classification, I presume this will be described in the
 > > setsockopt() manpages?  If so, wouldn't we need to be more precise in that
 > > manpage (can you provide the suggested text)?
 > 
 > The precise interaction between source address selection and the
 > SO_ALLZONES option is not defined by this case, and whatever man page
 > changes are made to document this will be explicit about this.

Sure.  However, by the way you trimmed my quoted text, I'm not sure if you
agreed to specify these:

 > We say "may" rather than "will" because other networking features --
 > such as IPsec policy and packet filtering -- may interfere, and the
 > implementation of this option is not required to punch through those
 > features.
 >
 > [ ... ]
 >
 > Not all protocols or protocol families will recognize this new option;
 > an implementation for IPv4 and IPv6 will be made first; other
 > protocols may implement it as needed.

... in the proposed manpage?  If these attributes are left unspecified, it
seems premature to me to classify SO_ALLZONES as Stable.

-- 
meem
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to