Alex, I do not think there is any problem for QUIC. IPv6 Next Header field only 
takes
a value for UDP; it does not take a value for QUIC. It is QUIC's job to make 
sure that
everything beyond the UDP header is aligned properly for its own purposes.

Fred

> -----Original Message-----
> From: Alexandre Petrescu [mailto:[email protected]]
> Sent: Friday, December 17, 2021 12:53 AM
> To: Templin (US), Fred L <[email protected]>; [email protected]
> Subject: [EXTERNAL] Re: [Int-area] Side meeting follow-up: What exact 
> features do we want from the Internet?
> 
> EXT email: be mindful of links/attachments.
> 
> 
> 
> 
> 
> Le 15/12/2021 à 19:56, Templin (US), Fred L a écrit :
> > Alex,
> >
> >> A new feature in an ALv2 would be that instead of just a
> >> frag-reassembly from sender to receiver, one would consider
> >> group-degroup of packets too, to reduce overhead.  If this too is
> >> not already there in ALs.
> >
> > The adaptation layer is that layer below the network layer but above
> > the data link layer. The group-degroup functions you are referring to
> > would apply at the adaptation layer send-side entry or receive-side
> > exit.
> 
> Yes, from a topology perspective that's where 'group-degroup' could sit.
> 
>  From an ISO layer perspective a 'group-degroup' would probably sit in an
> adaptation layer, below IP and above data link layer.
> 
> A 'group-degroup' function can also be seen in the jumbogram software
> mechanism, rather than in the frag-defrag mechanism.  The jumbogram
> seems to be in the app layer (ApL?) whereas frag-defrag in the
> adaptation layer (AdL?).
> 
> This makes wonder whether a group-degroup function would go into ApL or
> rather in the AdL?
> 
> Whether or not the modern QUIC, rather than the older UDP and TCP,
> includes a jumbogram mechanism and potentially group-degroup functions
> could be explored.  Because IPv6 jumbograms are specified for TCP and
> UDP (RFC2675 "jumbograms"), but not for QUIC.  This could further hint
> on which layer is best for a group-degroup function, AdL or ApL.
> 
> Alex
> 
>   Those
> > sorts of capabilities are already out in the wild in some systems,
> > but I have not seen a formal IETF spec yet. This seems like something
> > that could be mentioned in an adaptation layer spec.
> >
> > Fred

_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area

Reply via email to