The MPLS WG Archive

[Yangguang Xu <xuyg@lucent.com>]

ERO should contain the router ID for several reason:

1) Simplify receiving node operation and avoid the ambiguity as Vach and Bora
mentioned

2) Keep ERO operation consistent. (A node typically strip off all information of
itself before sending to next hop, so the outbound interface ID in ERO should be
stripped off)

3) Consistent format of unnumbered FA and unnumbered link. These two types of
links don't need to be differentiated especially if the unnumbered FA is through
an circuit switched transport network.

4) For CR-LDP, we don't need to invent another PHOP TLV

Yangguang

Vach Kompella wrote:
> 
> I apologize if this has been discussed ad nauseam.  I looked through the
> archives before venturing out on the subject, and couldn't find an
> (explicit) answer to the question: why isn't the router ID in the ERO, and
> how come 6.3 of the rsvp-unnum draft says:
>         "If, after processing and removing all initial subobjects in the ERO
> that refer to itself, the receiving node find a subobject of type Unnumbered
> Interface ID, it determines the next hop as follows..."
> 
> I can understand the router ID not being there makes it more concise, at the
> expense of being a little less readable, and requiring a little more
> processing at the receiving end.  In the interest of clarity and easier
> processing, I'd have eaten the 4 byte hit.  But that's something the WG can
> vote (has voted?) on.
> 
> The other aspect is the first clause of the quoted sentence above from the
> rsvp-unnum draft: "all initial subobjects that refer to itself."  Suppose
> the ERO is made up of Unnum links (1, 1, 1, 1, 1), i.e., a series of hops,
> each of which used their Interface #1 outbound.  Processing according to
> 6.2, a receiving node would strip out the first subobject, after validating
> that it received the ERO over its PHOP's interface #1.  Then, by the
> statement above, it looks at the next one which coincidentally appears to
> refer to itself also.  If I'm not missing something here, this first hop
> removes all the subobjects before it is done, which is not what is was
> intended.  With a router ID in each subobject, the ERO would look like ((r1,
> 1), (r2, 1), (r3, 1), (r4, 1), (r5, 1)), which would remove the ambiguity.
> 
> Can someone clarify?  Thanks.
> 
> -Vach