Re: IP Address in the ACM (Was: Comments onAccessControlModel- BNF)

[robert byrne <robert.byrne@Sun.COM>]

Kurt,

True.  The difference to me though is that "ipaddress" is a perfectly
well defined thing and so there is no a priori reason to make it
optional.  
This is not true of all subjects.  For example a "role" is still
undefined, so it is natural that some servers might not support it. 
Similarly for the case of SASL mechanisms--some implementations might
not have gotten around to implementing some of the mechanisms.  In these
cases there is nothing we can do.  However, by not making ipaddress
optional we are ensuring that we do not add, unecessarily, to these kind
of grey areas of the model.

Rob.

"Kurt D. Zeilenga" wrote:
> 
> Robert,
> 
> Given that the document allows additional subjects, there are
> already interoperability issues of the nature you seem to be
> concerned with.  If someone specified a "foo" subject to a
> server that doesn't support the "foo" subject, an error should
> be returned.  That ensures the client understands this server
> doesn't support the "foo" subject.  How is that different for
> the server returning a error for the IP-Address subject?
> 
> Kurt