Rich, Martin,

Please see inline.

On 11 May 2011, at 16:44, Richard Alimi wrote:

> Comment inline with most of the textual discussion omitted...
> 
> On Wed, May 11, 2011 at 1:44 AM, Martin Stiemerling
> <[email protected]> wrote:
> 
> ...
> 
>>> 
>>> I can see that an ALTO server may want to return different responses
>>> depending on the category of resource (or possibly more precisely the
>>> category of consumer) and we don't want to preclude that, e.g.
>>> returning a different response to a P2P client than it would to a CDN.
>> 
>> The assumption until now was really focused on the p2p use case and not on 
>> the CDN use case. However, I can see that in the CDN use case, an ALTO 
>> server may return a different response for CDN. For P2P on the other hand, 
>> we definitely do not want to identify the resource category.
>> 
>> Adding the possibility to identify the category of resource to the ALTO 
>> requirements would require a WG consensus IMHO.
>> 
>> The requirements would read along these lines:
>> (a) that the ALTO protocol MUST NOT require the specific resource to be 
>> accessed to be included in ALTO requests.
>> (b) that the ALTO protocol SHOULD allow to specify the resource to be 
>> accessed to be included in ALTO requests.
>> 
>> (a) rules out the requirement to have it in all requests, while (b) allows 
>> the protocol to have such a capability.
> 
> For the sake of completeness, I'll offer another one which was demoed
> (by Juniper) in the running code show. They allowed an ALTO Server to
> internally contain multiple "views" (read: set of maps), and it would
> serve maps based on the requesting client. The client could be
> determined by requesting IP address, supplied credentials, additional
> HTTP headers, etc.  This avoids us defining any additional
> identifiers, or classifying applications into any particular types in
> the protocol specification.
> 
> An early implementation of the P4P work used multiple views which were
> advertised by the server.  It was up to the client to determine
> whether they wished the default one, or wished to select a different
> one.

Before we go to far down the line of resource identifiers I think Martin may 
have misinterpreted what I meant slightly.

I was never thinking about including resource identifiers in ALTO. I was 
thinking of the multiple views scenario described above, e.g. an ALTO server 
"knows" (somehow) the client is a P2P client and gives it the P2P map which may 
expose different costs between the same endpoints (or even different sets of 
end points) to the CDN map for example.

I see know need to include categories of individual resources, but returning 
different maps based on the category of client/consumer (e.g. P2P client or CDN 
client, etc.) should not be precluded. Having said that I don't think that the 
category of resource client needs including in the protocol either.

I do think the original wording of the requirement could be improved though.

Apologies for sending you off on a tangent.

Ben

> 
> There are some reservations about including a resource identifier in
> ALTO.  In particular, we need to find a way to name it that allows
> this must work under different applications that exist today and in
> the future. There are ways to do it, but I'm just mentioning that this
> does open a can of worms (the DECADE WG is an example of this, and I'm
> sure there are others).
> 
> Yes, you proposed this as a SHOULD so we probably aren't bound to put
> this in the base protocol, but it would need to be worked out at some
> point.
> 
> Rich
> 
>> 
>>  Martin
>> 
>> 
>> [email protected]
>> 
>> NEC Laboratories Europe - Network Research Division
>> NEC Europe Limited | Registered Office: NEC House, 1 Victoria Road, London 
>> W3 6BL | Registered in England 2832014
>> _______________________________________________
>> 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