On Mon, 3 Feb 2003, Juergen Schoenwaelder wrote:
> Since forwarding architectures tend to look pretty different, I guess
> we can't do more than using some opaque identifier - whether that is
> an OBJECT IDENTIFIER, an INTEGER will some smart numbering scheme
> (similar to SnmpSecurityModel), or an OCTET STRING does not really
> make a big difference to me.

In principle any one of those could be made to work, yes.

> The problem is that mgmt applications either understand the
> semantics of a particular implementation or they will have to
> guess or ignore some rows.

I undestand that a management application may not have full
understanding of route table entries without understanding
the nature of the opaque identifier.  But I don't follow
why that implies that a management application has to ignore
some rows -- at least if all it's doing is inspecting rather
than modifying the rows.  The proposed index components are
destination prefix, opaque identifier, and next hop.  I can
still determine how many routes to a given destination via
a particular next hop exist (and query their other properties)
even if I don't know what the opaque identifier is.  [Note that
the order given for the indices may not be the optimal one --
probably the opaque identifier should be last -- but that does
not affect the information content.]

> So from an interoperability point of view, all this is only
> little better than not having such an index component.

I don't think I agree.  It's much better, because I can at least
represent a forwarding architecture that permits multiple routes to
a destination via the same next hop.  Without the opaque identifier
I could not do that.  More generally, without the opaque identifier
I could not map the table to an arbitrary forwarding architecture.  
It is true that it's best to split as much common stuff out of
the opaque identifier as possible.  If you can identify some
additional generic candidate index components beside destination
prefix and next hop please say so.

//cmh

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to