On 11 mrt 2008, at 14:38, Dave Thaler wrote:

> Note the specific clause that suggests that bridges can generate ICMP
> too big messages.

I've seen them do IP(v4) fragmentation, too.

>> On a link with variable MTUs the second step is not possible without
>> additional communication of per-destination MTUs, and even then
>> "invisible" layer 2 devices can get in the way. These layer 2 devices
>> can't do step 1 because of hardware limitations.

> I don't know what you mean about "can't do step 1".  Can you  
> elaborate?

Following topology:


H1----X----H2

In the current situation where X is a router, then X can by definition  
receive the largest packets that H1 can send because H1 and X MUST use  
the same MTU on their common subnet. If then the subnet towards H2  
uses a smaller MTU X can return a too big to H1.

For ethernet, the problem doesn't arise today because the MTU is  
always 1500 bytes.

If we now go to multi MTU subnets, H1 could send a packet that is  
larger than X can receive if H1 happens to have a larger MTU than X.  
This means that in that case X can't successfully receive the packet,  
which is of course a prerequisite for generating too big messages.
_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area

Reply via email to