Re: [ppsp] 答复: Selection of PPSP peer protocol draft

On Nov 29, 2011, at 7:58 AM, "Yingjie Gu(yingjie)" <guyingjie@huawei.com> wrote:

> I would like to recall the history of peer protocol. 
> We began to propose Tracker and Peer protocol at early 2010. Since the very beginning, we believe that Peer protocol should be consistent with Tracker Protocol. By consistence, I don't mean that they can co-exist, instead they should share the similar philosophy and should be able to interoperate.  That's why we call them a system, not two single protocols.
> 
> PPSP had spent a lot of time on discussion on whether the PPSP standards should use Tracker-based or Full-mesh architecture (finally, WG got consensus on tracker-based), and how the Tracker and Peer protocol should work, which is in the Requirements draft. We always regards the requirements draft as a guidance on Tracker and peer protocol. Based on the requirements draft, we proposed draft-gu-ppsp-tracker-protocol and peer-protocol. 
> 
> The WG has spent a lot of effort on PPSP Survey, because we want to extract common features on current PPSP system, instead of base on a single private protocol. 
> 
> I agree that Swift works, and it is a research project supported by many research groups, and I think those show support for swift are mostly involved in SWIFT project. 
> It will be easier for PPSP to get to an RFC published if we defining a protocol based on a mature private protocol. But what about other private protocols? ANY ONE of the protocols introduced in Survey draft can work well and most of them have much huge number of users than SWIFT, and even much better performance. Shall we take them into consideration?
> 
> I hate to make this selection as an attack on each other draft. And I hope those from SWIFT won't look at my email as an attack to their protocols. I just wonder what does PPSP WG want, a mature private protocol or a developing protocols based on widely survey? 

I would like to second Yingjie's question. It is important to discuss the objective first. There are indeed other options, e.g., based on reverse engineering, ppsp could use as a base one of those extremely widely deployed (with tens of millions of real users) protocols, as long as there is no IP issue.

Protocol design is not an optimization problem, i.e., pick the optimal protocol. So what is the goal of ppsp? If time is the essence and the goal is to have a peer protocol coming out of IETF quickly, then swift can be a starting base. Tanenbaum talked about the timing of standardization based on a camel shape. I feel that p2p streaming is already at the lower half of the body, i.e., many commercial protocols already exist. I am wondering on the opinions of the chairs and other seasoned IETFs on an approach that PPSP could take, to make a difference. I understand that many want to discuss the specific technical details, but then I see endless debates.

Thanks!

Richard

> 
> I agree with Cruz's opinion. This is not a selection of draft, we have more than one option. 
> 
> 
> 
> 
> 
> Best Regards
> Gu Yingjie
> 
> _______________________________________________
> ppsp mailing list
> ppsp@ietf.org
> https://www.ietf.org/mailman/listinfo/ppsp