Re: [p2pi] Information in an ALTO protocol

If the ALTO service provides ISP preferences, the whole issue of lying goes away.

If you say it's what you want, and it's really not, you'll be the only one unhappy.

Right incentives are the key.

-- Stas

-----Original Message-----
From: "stefano previdi" <sprevidi@cisco.com>
To: "Woundy, Richard" <Richard_Woundy@cable.comcast.com>; "Laird Popkin" <laird@pando.com>; "Salman Abdul Baset" <sa2086@columbia.edu>
Cc: p2pi@ietf.org
Sent: 9/12/08 9:14 AM
Subject: Re: [p2pi] Information in an ALTO protocol

Rich,

> From: "Woundy, Richard" <Richard_Woundy@cable.comcast.com>
> 
>> well, I don't think there's anything there that needs do be
> standardized
> 
> Sorry, I disagree. The protocol spec needs to make the semantics clear
> for a "user bandwidth" value.

sorry, I misunderstood. I wasn't referring to the "user bandwidth" here.

The point I was trying to make is that I'm not sure we need to standardize
the way rankings are computed since each provider may have own policies that
would make results to vary.

The choice of an ALTO client to follow ALTO "suggestion" (or not) will be
solely based on the effective performance.

> The spec could state that the value could
> be randomized / statistically balanced, so that such value manipulation
> could be anticipated by the ALTO client. Transparency is the key here.

if we have explicit BW values in the ALTO protocol, then I agree with you,
there's no point in hiding/cheating

> To be honest, I have my doubts whether randomization of the value will
> really help. 

"randomization" clearly not.

Sorry for the misunderstanding.

s.

> The more randomization is used, the less effective this
> information will help the ALTO client. The less randomization is used,
> the more concerns I have about user privacy.
> 
> I will step back from this debate and listen.
> 
> -- Rich
> 
> -----Original Message-----
> From: p2pi-bounces@ietf.org [mailto:p2pi-bounces@ietf.org] On Behalf Of
> stefano previdi
> Sent: Friday, September 12, 2008 9:22 AM
> To: Woundy, Richard; Laird Popkin; Salman Abdul Baset
> Cc: p2pi@ietf.org
> Subject: Re: [p2pi] Information in an ALTO protocol
> 
>> From: "Woundy, Richard" <Richard_Woundy@cable.comcast.com>
>> 
>>> Nothing prevents an ISP to configure his ALTO server in order answer
>> to
>> identical requests using some form of statistical-balancing
> preference.
>> 
>> I would worry that the 'statistical balancing preference' could
>> alternatively viewed as 'ISP ALTO server is lying' or 'ISP ALTO server
>> prefers some applications over others'. I don't want to go there.
> 
> well, I don't think there's anything there that needs do be standardized
> (or
> worked out in the context).
> 
> Regarding the Alto-lies, it looks to me ALTO is a form of traffic
> engineering based on peer selection guidance. I don't see why a
> dynamic/variable algorithm (taking into consideration parameters such as
> time of day, current backbone load, ...) returning different results for
> the
> same query, would be considered as a "bad thing".
> 
> But it also looks to me it's going to be something implementation
> specific.
> Being adopted/used or no will be the ISP choice.
> 
> s.
> 
>> 
>> -- Rich
>> 
>> -----Original Message-----
>> From: stefano previdi [mailto:sprevidi@cisco.com]
>> Sent: Friday, September 12, 2008 5:06 AM
>> To: Laird Popkin; Salman Abdul Baset
>> Cc: p2pi@ietf.org; Woundy, Richard
>> Subject: Re: [p2pi] Information in an ALTO protocol
>> 
>>> From: Laird Popkin <la
_______________________________________________
p2pi mailing list
p2pi@ietf.org
https://www.ietf.org/mailman/listinfo/p2pi