Re: [BEHAVE] Summary of the issues related to fragmentation and v4/v6 translation

[<eric.burgey at orange-ftgroup.com>]

Hello,
In fact in a NAT64 you need to wait for offset 0 fragment to be able to learn the appropriate  port number(source for outgoing packet, destination for incoming packets) and to be able to handle other fragments. Offset 0 fragment may be not the first received fragment. 
So probably the answer is more complex and is linked with the implementation. Some implementation may prefer to wait for all fragments of the same datagram, assemble the datagram, process the whole datagram in one operation and fragment the resulting datagram.
Some implementation may prefer to process packets as soon as possible and will wait only for offset 0 fragment to process all received fragment of the same datagram. In that case they need to implement a specific map to link the source address, the destination address, and the datagram ID to the appropriate port number for non offset 0 fragment.

	Eric BURGEY

-----Message d'origine-----
De : behave-bounces at ietf.org [mailto:behave-bounces at ietf.org] De la part de marcelo bagnulo braun
Envoyé : vendredi 30 octobre 2009 19:25
À : Simon Perreault
Cc : behave at ietf.org
Objet : Re: [BEHAVE] Summary of the issues related to fragmentation and v4/v6 translation

right
thanks, marcelo


Simon Perreault escribió:
> Adding to your question list...
>
> marcelo bagnulo braun wrote, on 2009-10-30 06:10:
>   
>> - Fragment: Add fragment header, copy fields from IPv4 header, also make
>> sure
>>  that the packet does not exceeds 1280 bytes.
>>     
>
> If the fragment exceeds 1280 bytes, do you just fragment the fragment by itself,
> or wait for the other fragments to arrive then fragment the reassembled packet
> as a whole?
>
> If you're going to do reassembly anyway you might as well fragment the packet as
> a whole. However that would need changing text in the stateless draft (section
> 2.1, first paragraph).
>
> Simon
>   

_______________________________________________
Behave mailing list
Behave at ietf.org
https://www.ietf.org/mailman/listinfo/behave