[dns-operations] PMTUD of .org servers

[Michael Graff]

Right but I personally suspect that few recursive resolvers live on a  
network with less than Ethernet mtu. I do think many live behind  
fragment dropping firewalls. Spending time on pmtu for udp thus seems  
like a net loss to me.

If only we could examine the broken side of any brokenness better.

--Michael


On Jun 20, 2009, at 9:52, Florian Weimer <fw at deneb.enyo.de> wrote:

> * Michael Graff:
>
>> Without and feedback on pmtu the local interface Mtu would be used I
>> think. That would normally result in a maximum sized packet per
>> fragment.
>
> It doesn't if the link MTU is lower.  The oversized fragments will be
> split further.  That's where the 5 (well, I wrote 7, but it should be
> 5) fragments come from.
>
> There is no reassembly in transit, unless you use stock Linux for
> forwarding, which you probably shouldn't (unless you know what your
> doing, usual disclaimers apply etc.).