On Thu, Feb 12, 2009 at 4:41 PM, stefano previdi <[email protected]> wrote:

>
> On Feb 12, 2009, at 4:32 PM, Saverio Mascolo wrote:
>
> ALTO aims at providing  informations on network status, i.e. feedback,  so
> that applications can make smarter decision.
>
>
> well, ALTO is delivering a ranking/preference system and not any
> information about network status.
>
> Now, network status _may_ be used in order to derive/compute a given
> preference. This is in hands of the ALTO provider and not supposed to be
> standardized in any form.
>

the use of the ranking/preference as a base for any decision will close the
loop. this may  trigger closed loop dynamics.  TCP is an example of closed
loop dynamics, even though in a much poor information context.

-s



> A potential warning here comes from the fact that the applications  close
> the loop and this maybe destabilizing.
>
> Thus,   standardiation should also involve how to use  ALTO informations to
> pursue  objectives such as load balancing routing etc. In other terms
> standardization should involve information and how to use it.
>
>
> I don't think I agree here.
>
> Standardizing "use of ALTO information" assumes you enforce a specific
> behavior. It will be unpractical and (IMHO) limited in its effectiveness.
>
> s.
>
>
>
>
>
>
> 2009/2/11 Stanislav Shalunov <[email protected]>
>
>> We haven't discussed anything about ALTO influencing actual underlay
>> routing or anything of the sort.
>> So I suppose it makes apps smarter by giving them more knowledge about the
>> network and the routing preferences.
>>
>>
>> On Wed, Feb 11, 2009 at 1:03 PM, DePriest, Greg (NBC Universal) <
>> [email protected]> wrote:
>>
>>>  Makes perfect sense to me.
>>>
>>>
>>> Is it fair to say an ALTO server adds intelligence to the network?
>>>
>>>
>>> Or is it more accurate to say that ALTO enables more intelligent app
>>> execution via the availability of network [and perhaps other] information?
>>>
>>>  ------------------------------
>>>
>>> *From:* Stanislav Shalunov [mailto:[email protected]]
>>> *Sent:* Wednesday, February 11, 2009 3:54 PM
>>> *To:* DePriest, Greg (NBC Universal)
>>>
>>> *Cc:* [email protected]
>>> *Subject:* Re: [alto] differences among applications
>>>
>>>
>>> That's, to me, the idea of ALTO.  The apps using information about ISP
>>> routing preferences and the network to improve peer selection.
>>>
>>> Note that this is a very broadly applicable technique: sure, BitTorrent
>>> and other P2P apps are most obvious users to begin with, but any sort of app
>>> that has a choice of network destinations can benefit.  Think along the
>>> lines of CDNs, HTTP mirrors, or DNS servers choice, for example.
>>>
>>>
>>> On Wed, Feb 11, 2009 at 12:11 PM, DePriest, Greg (NBC Universal) <
>>> [email protected]> wrote:
>>>
>>> Just to be sure:  You envision the app selecting peers for specific
>>> pieces of content and peer selection will use network data of some type in
>>> doing so?
>>>
>>>  ------------------------------
>>>
>>> *From:* [email protected] [mailto:[email protected]] *On Behalf
>>> Of *Stanislav Shalunov
>>> *Sent:* Wednesday, February 11, 2009 2:51 PM
>>>
>>>
>>> *To:* Zoran Despotovic
>>> *Cc:* [email protected]
>>> *Subject:* Re: [alto] differences among applications
>>>
>>>
>>> As others pointed out, an ALTO protocol is not expected to make peer
>>> selections for the apps.  On a high level, it's expected to provide
>>> information about the network and about ISP routing preferences.
>>>
>>>
>>> While peer selection preferences vary from application to application
>>> substantially, the network itself is the same, and so the information about
>>> it remains valid.
>>>
>>>
>>> -- Stas
>>>
>>>
>>> On Tue, Feb 10, 2009 at 2:14 AM, Zoran Despotovic <
>>> [email protected]> wrote:
>>>
>>> Hi all,
>>>
>>> I was wondering if and how IETF would address possible differences among
>>> relevant P2P applications in the sense that different applications may
>>> require totally different solutions. Was there any discussion on this before
>>> on the list?
>>>
>>> Just as an example, different criteria to drive peer selection may work
>>> differently for give-to-get streaming and tit-for-tat BT. So how will IETF
>>> deal with this? Standardize different solutions for different applications?
>>> Standardize one solution for all? Pick the most critical (heaviest traffic)
>>> applications and standardize a solution for it?
>>>
>>> It makes sense to clarify that at this early stage and, perhaps, first
>>> see if the solution should and can be application agnostic or not.
>>>
>>> Best regards,
>>> Zoran
>>>
>>> --
>>> Zoran Despotovic, Ph.D.
>>> Senior Researcher
>>>
>>> DOCOMO Communications Laboratories Europe GmbH
>>> Landsbergerstrasse 312, 80687 Munich, Germany
>>> Tel: +49-89-56824-205  Fax: +49-89-56824-300
>>> http://www.docomoeurolabs.de/
>>>
>>> Managing Directors (Geschaeftsfuehrer):
>>> Dr. Toru Otsu, Dr. Narumi Umeda, Mr. Tsutomu Sakai
>>> Amtsgericht Muenchen, HRB 132976
>>> --------------------------------------------------
>>>
>>> _______________________________________________
>>> alto mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/alto
>>>
>>>
>>>
>>>
>>> --
>>> Stanislav Shalunov
>>> BitTorrent Inc
>>> [email protected]
>>>
>>> personal: http://shlang.com
>>>
>>>
>>>
>>>
>>>
>
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to