[alto] Is Alto helpfull or harmfull ?

[Robert Raszuk <raszuk at juniper.net>]

Hi Paul & all,

I kind of share your opinion here.

What I would like to better understand reg Alto what is the value add 
such ISP domain scoped device would provide ?

P2P clients by themselves can measure today bandwith, delay, latency 
between content peers and adjust their peer selection accordingly at the 
application layer.

Those peers may cross ISP domain boundaries or may not. I think that 
actually there may be a little conflict of interests ... ISPs would like 
to save some chunk of inter-as bandwith for the cost of keeping the p2p 
volumes local while p2p protocol designers have been quite intelligent 
of making the peer selection natively in the application.

Perhaps there is a clear prove that the above does not hold anymore but 
if so I would like to hear about them so don't hesitate to reply pls :).

Cheers,
R.


> At 1:42 PM -0500 11/15/08, Laird Popkin wrote:
> > I think I see quite a few incorrect assumptions about ALTO that
> > you're making in this discussion.
>
> Um, no. I didn't make any of those assumptions; if I had, I would
> have stated them. People who know me will tell you that I err on the
> side of verbosity, for better or (often) worse.
>
> The assumption that I made, that I stated, is that the current large
> BitTorrent trackers would not want to use Alto information because
> they purposely do a bit of traffic management in order to spread the
> torrents out so that their communities are larger and more stable. I
> heard this directly from some folks who run trackers about two years
> ago.
>
> After I started this thread, I have hear that, since then, the world
> has changed. Many BitTorrent clients in the past year have started to
> do what can be described as "discovery of additional peers". That is,
> if a tracker tells Client A that its peers are B, C, and D, A will
> ask each of them who they are peering with and, if (for example) C
> says it is also peering with E, A will add E to its lits of peers. I
> have also now been told that the largest trackers like this because
> it reduces their traffic because they have to update their peer lists
> less often.
>
> Assuming that this new information is true (it is certainly
> plausible, given the scaling issues that the largest tracker is
> seeing), then my objection to the implications in the problem
> statement document is wrong. I'll do more research, but in the
> meantime, let's assume that the problem statement is correct and that
> Alto might have a positive effect on the current flood.
>
> --Paul Hoffman, Director --VPN Consortium 
> _______________________________________________ alto mailing list 
> alto at ietf.org https://www.ietf.org/mailman/listinfo/alto

_______________________________________________
alto mailing list
alto at ietf.org
https://www.ietf.org/mailman/listinfo/alto