Re: [v6ops] A good example of why we need to careful about ULAs

[Owen DeLong <[e-mail]>]

On May 31, 2013, at 9:07 PM, Brian E Carpenter <[e-mail]> wrote:

> On 01/06/2013 14:44, Owen DeLong wrote:
>> On May 31, 2013, at 7:21 PM, Ted Lemon <[e-mail]> wrote:
>> 
>>> On May 31, 2013, at 10:01 PM, Owen DeLong <[e-mail]> wrote:
>>>> How is having them leak in attempted TCP connections worse than having them leak in ICMP headers? It's still an invalid header.
>>> The ttl exceeded return packet doesn't expect a return packet of its own, so the fact that there's a "bogus" IP address on the header causes no harm, despite being technically wrong.   Filtering the packet based on the source address, on the other hand, would cause harm.   Nobody ever said life was going to be simple.
>> 
>> The TTL Exceeded packet shouldn't get forwarded by any router as a result of the LL Source address.
>> 
>> Thus it should never reach its destination.
> 
> Agreed. But when it does get through, it does no particular
> harm (you just see a bizarre hop in the traceroute). If it got
> through in a SYN/ACK exchange, it would probably distress a user.
> 

If it gets through, then it is an indication that every router between the one that sourced the packet and the recipient host is broken.

True, the packet itself is "mostly harmless", but…

>> 
>> Thus it will break traceroutes.
>> 
>> Thus it is harmful.
> 
> I'm not defending bad practice; just observing that it doesn't
> cause a disaster in trace route.

And I'm saying that if any of the routers between the LL-packet generating router and the recipient are working correctly, it does, in fact, break traceroute.

Owen