RE: About Teredo authentication indication

["Christian Huitema" <huitema@windows.microsoft.com>]

 > > When both the
> > client identifier and the authentication value are set to null, the
> > authentication indication gets a total length of 13 bytes. In such a
> > case, the encapsulated IPv6 packet will not start from a memory
> > address that is divisible by 4. Doesn't this cause problems on
> > platforms that need to worry about memory alignment?
> 
> Yes. With the current specification, it's necessary to move the packet
> in memory prior to reading/writing the Origin indication and/or the
> encapsulated IPv6 packet header.

The authentication header is only present: when authentication is actually used, in which case the client identifier and authentication values are not set to null; and when in the absence of actual client identification client and server use a nonce to secure the RS/RA exchange. In the second case, the alignment may be weird, but it only concerns a very small fraction of the packets; the regular data packets and the Teredo bubbles will be properly aligned. So, I don't think we have a really big problem...

-- Christian Huitema