Re: [Int-area] 回复： Request a WG adoption call for draft-xu-intarea-ip-in-udp

[Tom Herbert <[e-mail]>]

On Thu, May 17, 2018 at 7:18 AM, Behcet Sarikaya <[e-mail]> wrote:
>
>
> On Wed, May 16, 2018 at 10:22 PM, 徐小虎(义先) <[e-mail]>
> wrote:
>>
>> It doesn't matter whether or not it's already there. IMHO, given the
>> popularity of different overlay technologies such as VXLAN and MPLS-in-UDP
>> in practice, GUE initially and mainly targeted as a DC overlay approach has
>> little change to be widely deployed within data centers.
>>
>> As such, if the only possible applicability of GUE is for directly
>> carrying IP over UDP, I don't understand why we need such a overhead
>> associated with the variation of GUE. In another word, why not directly
>> assign a port to indicate IP-in-UDP, instead of using the GUE protocol
>> variant number to indicate. By the way, this the GUE protocol variant number
>> usage reminds me of the notorious misuse of the first nibble of the MPLS
>> payload to indicate the type of the MPLS payload:)
>>
>
> I agree and support the adoption.
>
> I supported GUE in the past.
> Why not have another way of UDP encapsulation with the possibility of a
> different area of applicability?
>
Applicability is one problem with this proposal. From the draft:

"This IP-in-UDP encapsulation technology MUST only be used within
networks which are well-managed by a service provider and MUST NOT be
used within the Internet."

That is incredibly limiting and unnecessary as other IP over IP
encapsulation protocols don't impose this sort of restriction. Also,
if the protocol specifies that it can only be used in "networks which
are well-managed by a service provider" then there needs to be a
normative description of what a "well-managed network" is. I suspect
this might have been motivated by GRE/UDP (RFC8086) that describes a
general Internet applicability scenario and one for networks that are
traffic controlled. The reason we needed to make this distinction is
because GRE can carry non-IP protocols for which we can't make any
assumptions about congestion control. For IP protocols it is assumed
that the protocols are properly congestion controlled, so for IP over
IP (like IP over UDP) there is no need have special considerations for
use over the Internet or in a traffic controlled network.

I would ask the authors of this draft to look closely at RFC8086 (and
RFC7510). Except for the aformentioned congestion control and a few
GRE specififc, RFC8086 addresses all of the common issues of IP over
UDP encapsulation including UDP checksum, fragmentation and MTU, ECMP,
diffserv, security, etc. The current IP over UDP draft doesn't
adequately address the issues and it would be far easier to leverage
all the work that went into RFC8086 than to redo the work here.

Tom



> Regards,
> Behcet
>>
>> Xiaohu
>>
>>
>> ------------------------------------------------------------------
>> From:Joe Touch <[e-mail]>
>> Send Time:2018年5月16日(星期三) 15:45
>> To:徐小虎(义先) <[e-mail]>
>> Cc:Tom Herbert <[e-mail]>; Internet Area <[e-mail]>;
>> intarea-chairs <[e-mail]>; draft-xu-intarea-ip-in-udp
>> <[e-mail]>
>> Subject:Re: [Int-area] 回复： Request a WG adoption call for
>> draft-xu-intarea-ip-in-udp
>>
>> It’s not complex. It’s already there. So there continues to be no reason
>> to waste either a port number or further time discussing this draft.
>>
>> Joe
>>
>> On May 15, 2018, at 9:01 PM, 徐小虎(义先) <[e-mail]> wrote:
>>
>> IMHO，there seems no need to introduce such complexity into GUE just for
>> the purpose of saving one port number.
>>
>> Xiaohu
>>
>>
>>
>>
>>
>> 来自钉钉专属商务邮箱
>> ------------------------------------------------------------------
>> 发件人：Tom Herbert<[e-mail]>
>> 日　期：2018年05月16日 11:55:49
>> 收件人：徐小虎(义先)<[e-mail]>
>> 抄　送：Erik Kline<[e-mail]>; Internet Area<[e-mail]>;
>> draft-xu-intarea-ip-in-udp<[e-mail]>;
>> intarea-chairs<[e-mail]>
>> 主　题：Re: [Int-area] Request a WG adoption call for
>> draft-xu-intarea-ip-in-udp
>>
>> On Tue, May 15, 2018 at 8:33 PM, 徐小虎(义先) <[e-mail]>
>> wrote:
>> > Hi Eric,
>> >
>> > Good question. This draft (draft-xu-intarea-ip-in-udp) describes a
>> > native
>> > UDP encapsulation scheme for IP packets, which is straightforward and
>> > light-weighted, just as MPLS-in-UDP [RFC7510] and TRILL-in-UDP
>> > (https://tools.ietf.org/html/draft-ietf-trill-over-ip-16#page-20) and
>> > etc.
>> >
>> GUE variant 1 implements native UDP encapsulation for IPv4 and IPv6.
>> Except for a different port number, there is no protocol difference
>> between that and doing IP in UDP as separate protocol.
>>
>> Tom
>>
>>
>> > Best regards,
>> > Xiaohu
>> >
>> > ------------------------------------------------------------------
>> > From:Erik Kline <[e-mail]>
>> > Send Time:2018年5月16日(星期三) 11:07
>> > To:徐小虎(义先) <[e-mail]>
>> > Cc:intarea-chairs <[e-mail]>;
>> > draft-xu-intarea-ip-in-udp <[e-mail]>;
>> > Internet Area <[e-mail]>
>> > Subject:Re: [Int-area] Request a WG adoption call for
>> > draft-xu-intarea-ip-in-udp
>> >
>> > Should this document make some comment about its relation, or lack of
>> > relation, to https://tools.ietf.org/html/draft-ietf-intarea-gue ?
>> > On Wed, 16 May 2018 at 11:53, 徐小虎(义先) <[e-mail]>
>> > wrote:
>> >
>> >> Hi co-chairs,
>> >
>> >> We would like to request a WG adoption call for this draft (
>> > https://tools.ietf.org/html/draft-xu-intarea-ip-in-udp-07) since it has
>> > been stable enough and the solution as described in this draft is needed
>> > in
>> > practice.
>> >
>> >> Best regards,
>> >> Xiaohu (on behalf of all co-authors)
>> >> _______________________________________________
>> >> Int-area mailing list
>> >> [e-mail]
>> >> https://www.ietf.org/mailman/listinfo/int-area
>> >
>> >
>> >
>> > _______________________________________________
>> > Int-area mailing list
>> > [e-mail]
>> > https://www.ietf.org/mailman/listinfo/int-area
>> >
>> _______________________________________________
>> Int-area mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/int-area
>>
>>
>> _______________________________________________
>> Int-area mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/int-area
>>
>
>
> _______________________________________________
> Int-area mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/int-area
>