On Sun, Nov 27, 2005 at 08:00:32PM -0800, Vishwas Manral wrote:
> Hi Fred,
> 
>  
> 
> Good point. I agree, however a bigger limit would provide more
> protection, besides a lot of extension headers may not be valid in most
> cases, so TCP headers would come within the 800 bytes. Having a
> configurable minimum value with default closer to 800, could help too.

Not quite sure what you mean, since at least one fragment might be
smaller, but I suppose you mean the initial. If we were to have a
minimum, I would suggest 640 (based on what I posted earlier in
this thread).

However, with IPv6 it is perfectly possible to add so many extension
headers that the TCP header is beyond whatever boundary you might
set. I think perhaps that's what Fred was saying. I suppose you
could just drop such packets assuming it is some kind of attack
though.

Another thing is that by doing filtering in end-hosts this would
be less of an issue. You can reassemble the entire packet before
making decisions.

> Pyda, on another note I have been wondering whether NAPT-PT work
> properly in the case where the first fragment, did not have the TCP port
> unless we maintained states of fragments (what the next header expected
> in the fragment is etc)?

Yes, it would make NAPT-PT difficult, but not so sure I want to
have NAPT-PT anyway...

Stig

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to