In my mind, this feature can be fulfill the function as an IPv4 address sharing scheme, treating some of the port number bits as part of an extended address. Perhaps PCP WG can consider it jointly with softwire WG.
Best Regards, Tina TSOU http://tinatsou.weebly.com/contact.html -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Dan Wing Sent: Tuesday, December 07, 2010 11:18 AM To: [email protected] Subject: [pcp] discuss: multiple port allocations This is one of the PCP discussion points. (This repeats a thread from earlier this week, but encloses it into the set of things we need to resolve so I can successful edit pcp-base-01.) http://tools.ietf.org/html/draft-wing-softwire-port-control-protocol-02 included the ability to request multiple ports. Requesting multiple mappings is an optimization to reduce PCP request/response messages. PCP is trying to stay as simple as possible to finish the specification soon and to keep implementations as simple as possible to meet the basic needs for port mapping -- and keeping it similar to how TCP and UDP create mappings in NATs (which is done one port at a time) goes a long ways towards that goal. I haven't found many protocols that would have much benefit from this optimization, though. Web servers need two ports (HTTP and HTTPS), IMAP and POP servers also need two ports (for their unencrypted and encrypted variants) -- but getting a webserver or mail server running is not time- or bandwidth-critical. RTP needs two ports (RTP+RTCP; separate issue if they need to be adjacent) and it is time critical. However, an application using RTP can pre-allocate its ports (at application startup) using PCP so this is not important, either. And the 4 PCP messages (to allocate two ports at application startup) are not significant compared to the RTP traffic (when a call is active) or to the application-level SIP (or Skype) keepalive traffic. Is there an application protocol that would benefit from the ability to get 2, 3 or 10 ports in one PCP message? If so, let's talk about that application protocol and understand how it could benefit. _______________________________________________ pcp mailing list [email protected] https://www.ietf.org/mailman/listinfo/pcp _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
