Richard, Stas,
On Mar 27, 2009, at 3:02 AM, Woundy, Richard wrote:
The WG chairs' diagram in the second ALTO session today portrayed a
war between ISPs unwilling to provide topology information to the
ALTO client, and P2P applications unwilling to provide any
information to the ALTO server.
The image is strong, but it does not seem to reflect the reality
that we are seeing. From our perspective, while the ISPs have
constraints about what they can disclose, the ISPs are still able
to provide useful policy information to the P2P applications. P2P
applications have concerns about the privacy of their users, but do
want to cooperate with the ISP to optimize network traffic in a
manner beneficial to ISPs.
that is good and it is also the impression we all had after the p2pi
workshop last year.
What confused me during the second meeting is that different
"approaches"
where presented as if there were to be XOR'ed. The impression was given
by the "need to find a good balance between the extremes" while in fact
the "extremes" of the first picture could easily coexist.
We acknowledge the fact that some information will/should/must not be
delivered by currently most popular p2p applications but this doesn't
mean the protocol must not even think about having them in its
specification (the same is valid for the opposite direction SP->p2p).
The main reason is that ALTO will also serve a few other cases, e.g.,
where disclosure of IP addresses is not an issue. There are many cases
where CDN applications _do_ share details about upper layer so that
an ALTO-like server gives optimal hint on selection mechanisms.
We can argue about the "% of coverage" of these CDNs compared to
"current most popular p2p apps" but I hope we have in mind a larger
scope of applications leveraging ALTO and not strictly focus on what is
mostly used today.
Anyway, after the second meeting and after looking at the different
presented drafts, it looks to me we mostly all agree that the protocol
is not intended to enforce any form of information disclosure but it's
just a container of information that will improve selection mechanisms
in different kind of application overlays.
Most of the proposals I've read include different forms and variants
of what conveyed by the protocol so I tend to believe we all mostly
agree and we have a significant common ground that allows us to move
on into the next step of the merge process.
thanks.
s.
We are working together to solve an important problem that ISPs and
the Internet have in general: the inefficiency of overlay routing,
and lack of policy information available in machine-readable form.
We have a proposal that we believe helps solve the problem to the
satisfaction of both communities.
Can we agree as a WG that there is no 'hemispherical conflict'?
-- Rich Woundy, Comcast
-- Stanislav Shalunov, BitTorrent
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
