On 11/4/13 10:06 AM, "Wendy Roome" <[email protected]> wrote:

>I think we've already lost that battle.  First, the client has to
>understand HTTP 1.1, including features like chunking.  Second, JSON is
>simpler than XML, but it's not trivial to parse.

We were behind changing from XML to JSON. So, this was a conscious choice.
The original protocol was XML based.

>Third, the client really
>should understand the IRD to select the default network map and associated
>cost map resources, because the server may change the resource uris at any
>time. And finally, the client should detect duplicate CIDRs.


This we should *definitely* fix. If you/folks are at Vancouver we should
sit down and try to hash this out. We should have a _always work_, URI,
_fall back_ default network and cost map.


>
>If you really want to simplify things for a sensor, then some other ALTO
>client can retrieve the network map from the ALTO server, eliminate
>duplicates, recast the network map as (CIDR,PID) pairs, sort by increasing
>CIDR length, write the result into a plain text file in some web server,
>and give the uri to the simple client. The simple client reads the
>(CIDR,PID) pairs and matches against the CIDRs. First match gives the PID.
>
>       - Wendy Roome
>
>On 11/04/2013 12:37, "Reinaldo Penno (repenno)" <[email protected]> wrote:
>
>>ALTO client should be dead simple. In the very beginning of ALTO protocol
>>Bittorrent folks mentioned that a ALTO client should be able to work by
>>simply downloading a text file from a well known URL/IP address and doing
>>longest prefix matching so that it could be implemented in the simple
>>devices like sensors. This is something I still believe we should strive
>>for.
>
>

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

Reply via email to