RE: encoding link id for unnumbered interfaces

["Dovolsky, Dan" <ddovolsky@movaz.com>]

BSD.

Question:

Original definition of Local/Remote Interface IP Address sub-TLV, specified in draft-katz-yeung-ospf-traffic-02.txt, has divided into it's redefinition in draft-katz-yeung-ospf-traffic-06.txt for numbered i/f plus new Link Local/Remote Identifier sub-TLV in draft-ietf-ccamp-ospf-gmpls-extensions for unnumbered i/f.
Original 02 version covered both numbered and unnumbered cases.

If I understand it right, it was caused by not proper definition of unnumbered case in 02 version. In this case, why just not to fix this definition in 02?

Now, with katz-yeung and ccamp-ospf, implementation has to find link identifier in different sub-TLV, depended what type of interface (numbered/unnumbered) it processed.

Proposal:

Instead, we could have just one pair of Local/Remote Link Identifier sub-TLVs. For numbered i/f it will be Local/Remote IP Address and for unnumbered i/f it will be Local/Remote Link identifier 32 bits length as well. Sub-TLV type code the same as for Interface IP Address sub-TLV type.

Dan Dovolsky.

-----Original Message-----
From: Yakov Rekhter [mailto:yakov@juniper.net]
Sent: Friday, April 26, 2002 4:46 PM
To: Naidu, Venkata
Cc: ccamp@ops.ietf.org
Subject: Re: encoding link id for unnumbered interfaces 


> Yakov,
> 
> -> I'd like to suggest that we replace Link Local Identifier sub-TLV
> -> and Link Remote Identifier sub-TLV (in both ISIS and OSPF) with a
> -> single sub-TLV that would contain both Link Local Identifier and Link 
> -> Remote Identifier.
> 
> -> Any objections ?
> 
>   I prefer : different sub-TLVs. That gives consistency and
>              easy modification in future.
> 
>   I suggest: Rather make Remote Identifier *mandatory* when
>              Local Identifier is present. This is definitely
>              bad but little better than your suggestion. :-)
> 
>   I like   : Make both Local and Remote Identifiers are optional
>              (just like local and remote IP address TLVs in TE)
>              and leave the decision to individual vendors 
>              (local decision). 
> 
>   Is there any scenario where this causes interop issues?!?!
>   Is that you are worried about *Tag, Length* extra memory
>   consumption?

No, it is just that if you have both local and remote identifiers
in the same sub-TLV, you don't have to worry about such error
cases as the situation where you have remote identifier sub-TLV
present, but the local identifier sub-TLV missing.

Yakov.

P.S. In contrast with numbered interfaces that allow for both
p2p and multiaccess interfaces, unnumbered interfaces are
restricted to p2p only.