On Apr 25, 2008, at 12:39 AM, jiangxingfeng 36340 wrote: > Hi, all: > > IMO, another important work is to make clear the assumptions of the > future work, such as: > 1. Which kinds of data will be stored in the overlay? what's the > reasonable size of the data?
The only reasonable answers are: For the protocol: any and any. For a particular instance: whatever policy allows. > > > 2. How much churn of the overlay where the peer protocol will be > applied? The pratical way is to use the experimental data from the > real existing system. I am not sure the resulting peer protcol could > work in all cases. I don't think that is productive. We can't predict what churn is realistic, given that I can create networks with close to zero churn and networks with node lifetimes measured in seconds. As usual, performance and data availability will degrade as churn increases. > > > 3. How much percent of the public nodes in the overlay is reasonable? Anywhere from 0 to 100%. Again, a protocol design needs to be robust in these parameters; we're not designing a specific instance, but rather something that works across the Internet, present and future. Clearly, there will be configurations where nothing will work and things will work faster in 'nice' configurations, but that's about all we can say. Hopefully, our design will work as well as it can in any given circumstance. > > > IMHO, if we don't work it out before we come up with the protocol, I > think, it's hard to get consensus on the proposal, because we have > no common assumptions. I'm sorry to say that if this is required, we may as well fold now. > > > Comments? > > Regards > JiangXingFeng > > _______________________________________________ P2PSIP mailing list [email protected] https://www.ietf.org/mailman/listinfo/p2psip
