In addition to legitimate concern with the economics of IETF proposals
and standards, I would encourage the group to also be concerned with how
those standards will be used.

"...when legal shifts to P2P" certainly acknowledges how most P2P is
used today and is likely to be used tomorrow (percentage wise) absent
constructive action by key stakeholders such as IETF. 

Due to a lack of such action, the concept of a "wild west" Internet has
driven a number of countries to begin to take steps to "tame the west."
New Zealand is the latest.

A better approach would be for IETF to consider the total breadth of the
"congestion" problem and consider comprehensive solutions addressing
engineering, economics, and law.

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of
Nicholas Weaver
Sent: Monday, March 09, 2009 11:48 AM
To: Wes Felter
Cc: Nicholas Weaver; [email protected]
Subject: Re: [alto] Comments on draft-weaver-alto-edge-caches-00


On Mar 8, 2009, at 4:09 PM, Wes Felter wrote:

> On Mar 8, 2009, at 12:15 AM, Nicholas Weaver wrote:
>>
>> But the key is "when legal shifts to P2P" (eg, like CNN's live  
>> video streaming), they must be able to cache it, because this is  
>> obviously coming down the pike, and spells bad news if it happens  
>> without caching.
>
> I see your point here. But this makes me wonder if P2P caching is  
> just a way for ISPs to subsidize P2P CDN companies -- another form  
> of cost shifting. (Will P2P CDNs bill content providers for total  
> bits delivered, even if those bits were delivered from an ISP-owned  
> cache?) I suspect economics of this sort is off-topic for IETF so I  
> won't ramble on any more.

Actually, I don't think its off topic, economics MATTER when building  
networks, especially P2P.  ISPs understand their economics well, but  
many of the P2P boosters don't understand this or refuse to  
acknowledge this.  Thus I'll argue that this is perfectly on-topic, as  
we want to do economically-relevant engineering.


As for caches themselves, if anything, I'd hope the ISP sees their  
interest is "don't bill".

Because:

Edge caches for P2P are cheap:  By dropping the reliability concerns  
(well, by shifting reliability concerns onto clients which have  
already addressed them), they cost almost nothing once the software is  
designed, and most of the software is already avaliable (since most is  
common with client code).

Thus the maximum benefit to the ISP occurs when they are maximally  
used.  Thus they don't need to build up the very extensive  
infrastructure that Akamai or another edge-located CDN does to  
tolerate failures.


If the edge caches aren't "free" to content providers, content  
providers simply won't use them, because they are shifting towards P2P  
because it is "free" to them as long as customers have "not quite  
unlimited" network connections.  EG, the norwegian public broadcaster  
has had a huge success in using BitTorrent in reducing their bandwidth  
bills.  Once a content provider takes the jump to full P2P, going back  
to a world where they have to pay MORE is not going to happen  
voluntarily.



Edge caches are partial-deployable:  If only ONE major ISP uses such  
edge caches, that ISP benefits and the content providers benefit, but  
other ISPs don't benefit or really suffer (assuming you have  
localization).  But partial-deployability is less effective in the  
case of charging for data from the cache, as you need to bootstrap up  
to at least a certain critical mass before you can bill.


Thus I think its one of those cases where IF the ISPs can resist  
charging for edge-cache data delivery, everybody involved in the  
content delivery process: the providers, the users, and the ISP, all  
win, with only the conventional CDNs losing.

_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to