On Fri, Jun 17, 2016 at 12:48 AM, Xuxiaohu <[email protected]> wrote: > > >> -----Original Message----- >> From: Joe Touch [mailto:[email protected]] >> Sent: Friday, June 17, 2016 2:57 PM >> To: Xuxiaohu; Tom Herbert >> Cc: [email protected]; [email protected] >> Subject: Re: [nvo3] [Int-area] Fwd: New Version Notification for >> draft-ietf-nvo3-gue-03.txt >> >> >> >> On 6/16/2016 11:41 PM, Xuxiaohu wrote: >> >> More than that, GUE was accepted as a WG doc *and* has already been >> >> assigned a port number. >> > Oh, a WG doc? a doc which has nothing to do with multi-tenancy but happens >> to be adopted by a WG working on multi-tenancy? >> I'm not advocating where this doc *should be* - or should have been - >> adopted. >> I'm simply noting that it already has been adopted. Which does carry weight >> in >> the IANA assignment of ports (as noted in RFC 6335). >> >> > >> >>> To save a port number, the header format is made ugly. Is it >> >>> worthwhile? If >> >> UDP port resource was so sparse as you had imagined, I think the UDP >> >> port resource keeper would not allocate two different port numbers >> >> for VXLAN and VXLAN-GPE since the P-bit in VXLAN-GPE header is enough >> >> to distinguish VXLAN-GPE from VXLAN. For more details, please look at >> >> section 3.2 of >> (https://tools.ietf.org/html/draft-ietf-nvo3-vxlan-gpe-02#page-6). >> >> VXLAN was assigned in 2011. >> >> >> >> VXLAN-GPE was assigned this year (2016). >> >> >> >> If what you say is correct*, then you might be correct in assuming >> >> that a VXLAN-GPE assignment might inhibit a later VXLAN assignment, >> >> but that's not the order things happened. >> > Your logic seems confused to me. My point is VXLAN-GPE should share the >> same port number (i.e., 4789) with VXLAN if the port number resource was so >> sparse. Unless that assumption is fake. >> >> Your logic fails to consider that these two requests were not made at the >> same >> time. Also, VXLAN was not made *after* VXLAN-GPE. If either of these were >> true, then the argument for a single port number would be important. > > More confused to me:) Let me make it simpler, VXLAN was allocated a port > number in 2011 which is 4789. VXLAN-GPE asked for a port in 2016, why > allocate a new number rather than reusing the port number 4789 provided the > port number resource was so sparse? Your answers to the above question seems > to be: 1)these two requests were not made at the same time, 2) VXLAN was not > made *after* VXLAN-GPE. For the first answer, did you mean, for two proposals > which could have shared the same port number, as long as they requests at > different time no matter intended or not, they would be assigned two > different port numbers. For the second answer, have you seen protocol X be > made after an extension protocol to X?:) > VXLAN-GPE requires a different protocol number, the P-bit was not sufficient. The problem was that VXLAN defined unknown flag bits to be ignored upon receive. So if a legacy VLXAN device ever received a VXLAN-GPE packet (P-bit set) it would be misinterpreted as a VLXAN packet. Effectively this makes VXLAN-GPE a new protocol not a different version of VXLAN. All of this was discussed on the nvo3 list.
Tom > Xiaohu > >> So, in brief, IMO (with my ports hat off) if you had a stronger argument for >> UDP-in-IP (i.e., you convinced a WG to adopt it) *and* you proposed it before >> GUE made its request, then things might have turned out differently. > > > >> Joe >> > _______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
