The MPLS WG Archive

["chetan kumar s" <chetan.kumar@wipro.com>]

On Tue, 19 Nov 2002, David Charlap wrote:

> chetan kumar s wrote:
> >
> > 	The session attribute object as a flag 'SE Style desired', which
> > mandates the egress to respond the resv message with SE style. Now if the
> > Path refersh messages have conflicting flags, how is this handled. I was
> > looking at the classical RSVP mailing list archive, found an interesting
> > thread on this:
> > http://www.cs-ipv6.lancs.ac.uk/ipv6/mail-archive/rsvp/1996-04/0068.html
>
> You have one critical mistaken assumption here.  The flags in
> SESSION_ATTRIBUTES are not mandatory.  They are advisory.  An egress

Yea, the session_attribute flags are not mandatory, but however, if the
flags are present(and set) the egress node SHOULD use SE style, the RFC
says. So it may not be good for egress to ignore.

The SE style desired flag is required in the session attribute so that
ingress can control the reservation style, since it is the ingress node
that decides if there is requirement for rerouting of tunnel without
tearing it down. Now if the egress node ignore any such request, one can
not expect consistent network operations..

Thanks
Chetan S

> node is free to ignore them if it wants to.
>
> This should also answer your second question.  If the ingress signals a
> Path requesting SE style, and an SE style Resv is returned, and then the
> ingress refreshes that LSP (or perhaps creates another LSP in the same
> session) without the SE-style request, the egress node should simply
> ignore the change and continue sending back SE-style Resv messages.
>
> Now, if the egress node sends back two different Resv messages with two
> different styles, then the node detecting that conflict should generate
> a ResvErr message.  But this shouldn't happen unless the egress node is
> broken.  A style-conflict error should only happen in a multicast
> situation, where two different egress nodes for the same session choose
> different styles.
>
> -- David
>