Re: [P2PSIP] draft-maenpaa-p2psip-self-tuning-01

[Jouni Mäenpää <[e-mail]>]

Hi Cullen,

Thanks for your comments. See inline for my answers.

Regards,
Jouni

Cullen Jennings wrote:

> I'm not sure what anyone else things, but I think it would be worth
> considering moving just one small part of to the base draft - namely
> Section 5.6.1.

I see no problem with doing that.

> On the Pro side for doing this:  It's really easy to implement and it
> would provide some interesting options for future things such as this
> self tuning draft.

The nice thing about self-tuning is that it automatically determines
appropriate values for the DHT parameters that are otherwise very
difficult to configure manually. An example is the stabilization rate.

> On the Con side for doing this: in many DHT, a large percentage of nodes
> leave without sending nay signaling that they are going to Leave so
> specifying what goes in a Leave message might be a little optimistic

I agree that many nodes may leave without sending any Leave messages
(i.e., crash). The mechanism works even without the data included in
Leave messages; including data in them is just an optimization that
speeds up recovery from peer departures in those cases in which peers
send explicit leave signaling. Section 6.3.1 of the draft deals with
failure detection (e.g., how for instance crashed peers are detected).