Hi Wendy,
On Wed, Jul 24, 2013 at 1:04 PM, Wendy Roome <[email protected]>wrote:
> Here's a new question for you: should an ALTO server that only offers the
> endpoint cost and properties services be considered a legitimate ALTO
> server?
>
> I think the protocol currently requires an ALTO server to provide a full
> network map and full cost map service, so an endpoint-only server is
> illegal.
>
> However, I think we should allow them. Why? Because that's what clients
> want. Clients really care about costs between endpoints. Clients do not
> care about PIDs.
>
> So why have PIDs and full cost maps at all? PIDs are necessary to make the
> endpoint space manageable. But PIDs are not an end to themselves, just a
> means to an end. Full cost maps exist for high performance clients, like
> busy p2p trackers, who might need to do 100's of queries per second.
>
> But for low-volume clients, an endpoint cost service is perfectly
> adequate, and much easier to use.
>
> Incidentally, a provider could work around the network map requirement by
> defining a trivial network map with one default PID (for 0.0.0.0/0) and a
> dummy cost map that returns no costs. But should we force a server to
> resort to that subterfuge?
>
> In terms of the protocol document, I think that just means dropping the
> "MUST provide" sentences from {10.1.1} and {10.1.2}.
>
>
I agree with this point. As my understanding of the potential deployment of
ALTO evolves, I see more diverse deployment scenarios. The case of only an
endpoint property service (e.g., to help with DC localization with a simple
mechanism), or only an endpoint cost service (proxidor), as you nicely
argued for, appear to be legitimate deployment scenarios for me, and adding
"dummy" maps to satisfy the spec to be conforming appears to be less
elegant/pure. How about we listen others' feedback, and then make a
decision in the next few days?
Thanks!
Richard
> - Wendy Roome
>
>
> _______________________________________________
> alto mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/alto
>
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
