Re: FW: OSPF WG Charter Proposal

[Jeff Parker <[e-mail]>]

I can't speak for the others, but here are my thoughts.

In OSPF, there is no restriction on what a PDU contains,
and a router can deal out an arbitrary hand from the
LSA in the DB.  IS-IS, on the other hand, fixes the
contents of a Link State PDU (LSP) at creation time.
If one TLV changes in IS-IS, we cannot simply flood
the updated TLV: we have to regenerate and reflood
the whole LSP.

This is all to say that the standard unit of exchange
in ISIS is a big honking LSP of 1492 or more bytes.
We decided to split this into TLVs (rather than hand
back a whole LSP) due to concerns about SNMP limits,
not because we thought is was the best thing to do.

This is moot in OSPF, as there is no "standard unit"
bigger than an LSA, and thus I wouldn't expect the
same issue to arise.

- jeff parker


> Hi Rohit,
>
> I was part of the "others" that Don mentioned in his other mail.
>
> Jeff P. correct me(I may have interpreted it wrongly), but I
> think we chose
> the TLV approach in ISIS as we could have probs because of "small SNMP
> buffer size and the lack of a defined way to fragment
> payload". In short an
> SNMP agent may not be able to fulfill the request for an LSP of large
> size(255 bytes??). In OSPF we anyway send the entire LSA as
> OCTET string.
>
> If we are not interpreting the TLV's, we could as well send
> the entire LSA
> itself as OCTET string. An OSPF-v2 TE LSA by the way would
> contain just one
> top-level TLV anyway. We could have a seperate table for TE
> LSA's though in
> this case.
>
> Thanks,
> Vishwas