Re: [MSDP] Reminder -- MSDP WG last call ends at 1700 PST Today

On Mon, 2 Dec 2002, David Meyer wrote:
> 	So please get comments to the list if you have them.

Ok, thanks fot the reminder -- forgot already.

As a novice in both MSDP and PIM-SM draft reading there were a few things 
that struct as quite confusing.

 1) (a bit besides the point here, but..) pim-sm-new draft introduction
text is as if every sender/receiver was in the same PIM domain.  Should
MSDP spec somehow clearly point out the differences in the model described
in the introduction, or even rewrite relevant parts of the introduction in
the MSDP spec?

 2) The bit:

--8<--
9. Intermediate MSDP Peers

   Intermediate MSDP speakers do not originate periodic SA messages on
   behalf of sources in other domains. In general, an RP MUST only
   originate an SA for a source which would register to it, and ONLY RPs
   may originate SA messages.
--8<--

I have hard time parsing this one, having little background in the failure
modes or wrong behaviour that's being referred to here.  If this means
what I think it means, one might add something like:

   Obviously, intermediate MSDP speakers still forward SA messages 
   received from other domains.

Or otherwise clarify what those "SA messages behalf of sources in other 
domains" actually are, I'm having difficulties figuring that out..

3) what the heck is this:

--8<--
15. Packet Formats

   MSDP messages will be encoded in TLV format. If an implementation
   receives a TLV that has length that is longer than expected, the TLV
   SHOULD be accepted. Any additional data SHOULD be ignored and the  
	MSDP session should not be reset.
--8<--

This can be interpreted two ways:

 * at first read, I thought it was like "(final) TLV has more data than 
Length specifies", apparently that's not the case, and for that the below 
applies:

I fail to see any reason (except coping with some possibly broken 
implementations) why wrong length should not be used -- if they have 
additional data to send, why don't they just increase the damn _Length_ of 
TLV's ??
 
 * at second read, I noticed that it probably means that length values
longer than 9192 should also be accepted (but the rest of the packet after
9192 bytes ignored).  This is a bit dubious too, but acceptable to ensure
interoperation, I guess.  If this is the case, perhaps:

 If an implementation
   receives a TLV that has length that is longer than expected, the TLV
   SHOULD be accepted.  Any additional data SHOULD be ignored and the   
   MSDP session should not be reset.

should be reworded like:

 If an implementation
   receives a TLV whose Length exceeds the maximum TLV length, 
   specified below, the TLV SHOULD be accepted.  Any additional data, 
   including possible next TLV's in the same message, SHOULD be ignored 
   and the MSDP session should not be reset.

4) I believe you should add some notion of MSDP SA rate-limiting features,
mentioned earlier in the draft, in the security considerations section;
it's just so easy denial-of-service attack, and people should be aware of 
that.

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords