Re: [ppsp] I-D Action: draft-ietf-ppsp-base-tracker-protocol-00.txt

Hi Arno and All,

As a matter of fact, in page 28 (second paragraph: implementation note) such situation is described, i.e., the Tracker service responds with a random sample of the peer population in the peer list.

The text however, as it is written, is not clear, "apparently" restricting the "random" nature of the sample taken to a non-specific selection criteria in the PeerNum element. That situation can perfectly be cleared by rewriting that paragraph.

What I mean is: the tracker protocol semantics already conveys the "criteria" to be respected by the Tracker service in the corresponding responses regarding the peer list, and the "normal" behavior of the Tracker Service should be of taking a random sample of the peer population in a swarm (subject or not to more fine-grained criteria expressed in the PeerNum attributes or by Operators’ policy preferences, e.g., peer location). 
As such, the peer protocol requirement 13.2.3 was already considered, but not conveniently expressed in the text.

Of course, as in João Pedro Taveira's response, an extension to the base protocol semantics could also be considered in order to include additional attributes in the PeerNum element to "instruct" the Tracker service of a "selector algorithm" to be used in addition to the other criteria, and in alternative to the default "random" selector. But that is an option for extension of the base protocol. 

Regards,

Rui Cruz
rui.cruz@ieee.org

IST/INESC-ID/INOV - Lisbon, Portugal
__________________________________________
ppsp mailing list
ppsp@ietf.org
https://www.ietf.org/mailman/listinfo/ppsp

On 19/02/2013, at 10:59, Arno Bakker <arno@cs.vu.nl> wrote:

> Hi all
> 
> good that the tracker protocol is now in a WG document. Reading it again I really would like to have an option to specify that I want the tracker to return a random sample of the peer population of a swarm. Rui previously argued (mail dd. 13-11-2012) that it is an implementation issue, but I think users must be able to ask for specific semantics, just like they can select the number of peers to be returned, etc.
> 
> What do you think?
> 
> CU,
>    Arno
> 
> _______________________________________________
> ppsp mailing list
> ppsp@ietf.org
> https://www.ietf.org/mailman/listinfo/ppsp