Re: Multi-AS needs : draft-zhang-mpls-interas-te-req-02.txt

Hi David,

At 10:01 10/03/2003 -0500, David Charlap wrote:
>Kenji Kumaki wrote:
>>I think the latter is correct.
>>We need to extend a existing protocol(e.g. RSVP-TE) because a current
>>RSVP-TE can't cover an inter-AS TE.
>
>What makes you say that?
>
>I don't think current implementations can support this.  But RFC 3209 
>includes inter-AS provisions.
>
>See section 7.4.
>
>One of the EXPLICIT_ROUTE subtypes is an AS number.  If one of these is 
>present in an EXPLICIT_ROUTE object, then the abstract node represented is 
>the entirity of the specified ASN.
>
>Implementations should treat this in a similar fashion to how they treat 
>an IP subnet ERO subobejct.  If it's the first subobject, then the router 
>must be in that ASN.  If it's the subobject corresponding to the next-hop, 
>then it should use that ASN to determine the next hop along the path to 
>that ASN.  If the node is strict, then the next-hop address should be a 
>part of that ASN.

This is just one of the aspects of inter-AS, there are many others listed 
in the inter-AS requirements drafts. The point that Kenji wanted to make 
was that the current set of protocols for intra-AS TE does not allow to 
address all the requirements listed in the requirements drafts. I just 
dropped on the TE WG mailing list another email mentioning some of the 
required protocol extensions.

>I'm not sure if BGP provides sufficient information for generating a 
>topology of ASNs (needed for generating an AS-hop-path).  An extension of 
>some form might be necessary for this to work.  But that would be a 
>routing extension, not an RSVP extension.  And it would only be necessary 
>if you want to dynamically generate inter-AS explicit routes - a feature 
>of dubious utility to begin with.

Similarly to the inter-area TE case, there are several possible solutions 
there: we proposed some two of them in the following 
drat:draft-vasseur-inter-as-te-00.txt

Thanks.

JP.

>-- David