Hi Tom, Fair enough. Just so I am not mis-interpreted, AERO is really agnostic when it comes to the encapsulation type. It can use its own encapsulation type, or it can use GUE, or it can use native IP-in-UDP, or it can use a minimal encapsulation like GRE, IP-in-IP, etc.
The facilities you cited below (fragmentation, security etc.) are a good reason to go with something like GUE. And, I agree the main advantage of IP-in-UDP (perhaps the only advantage?) is the 4 bytes per packet savings. Thanks – Fred [email protected] From: Tom Herbert [mailto:[email protected]] Sent: Monday, April 27, 2015 1:53 PM To: Templin, Fred L Cc: Brian Haberman; [email protected] Subject: Re: [Int-area] Why combine IP-in-UDP with GUE? On Mon, Apr 27, 2015 at 10:18 AM, Templin, Fred L <[email protected]<mailto:[email protected]>> wrote: > -----Original Message----- > From: Tom Herbert [mailto:[email protected]<mailto:[email protected]>] > Sent: Monday, April 27, 2015 10:10 AM > To: Templin, Fred L > Cc: Brian Haberman; [email protected]<mailto:[email protected]> > Subject: Re: [Int-area] Why combine IP-in-UDP with GUE? > > On Mon, Apr 27, 2015 at 9:41 AM, Templin, Fred L > <[email protected]<mailto:[email protected]>> wrote: > > In case anyone is wondering why I have suggested combining IP-in-UDP with > > GUE, there may be some uses where only some packets in a flow need to > > include the GUE header whereas the vast majority of packets could go as > > IPv4 or IPv6 raw encapsulation. So, having everything together under the > > same UDP port number could be advantageous. At least that's what I did > > in AERO. > > > What exactly would be the advantages of this? For one thing, it allows a natural separation of control plane and data plane (data plane as native IP-in-UDP; control plane as GUE). For another, it takes care of fragmentation using GUE encapsulation while unfragmented packets can go as native. This all comes at a savings of 4bytes per packet, which is debatable as to whether it is worth the trouble. But, if you think the overhead savings is insubstantial, I think you would probably not be in favor of IP-in-UDP native format whether/not it were bundled with GUE, right? It's just that I don't see much benefit in these approaches other than the four bytes savings. IMO, the main drawback of directly encapsulating IP in UDP is that it instantly becomes a feature frozen in time. We can never improve upon it or extend it-- we can't change IP fragmentation, we can't add a header checksum to get to circumvent the unpleasantness of using the UDP zero checksum with IPv6, we can't add security, etc. Thanks - Fred [email protected]<mailto:[email protected]> > > Thanks - Fred > > [email protected]<mailto:[email protected]> > > > > _______________________________________________ > > Int-area mailing list > > [email protected]<mailto:[email protected]> > > https://www.ietf.org/mailman/listinfo/int-area
_______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
