Re: request-routing requirements

[John Martin <jmartin@netapp.com>]

Here's my first cut. I apologise if I am not using the exact correct 
terminology but I hope this is clear nonetheless.

1. A Request Routing System (RRS) SHOULD (MUST?) know that the surrogate to 
which it directs a request for content is capable of delivering that 
requested content to that user.

2. Any authoratative element in a RRS SHOULD be able to answer a request 
for RR resolution though it MAY redirect the request to an alternative 
element. [I'm thinking of e.g. forcing DNS iterative versus recursive]

3. A RRS MAY be make a routing decision based on characteristics of the 
CONTENT, the various CDNs or other characteristics than the location of the 
requesting user and their proximity to a particular surrogate (e.g. the 
capabilities of that users agent, system load, preferred language).

4. A RRS SHOULD produce a resolution which is "transparent" to the end user.

...I'll try to think of some more.

John
---------------------------------------------------------------
Network Appliance           Direct / Voicemail: +31 23 567 9615
Kruisweg 799                               Fax: +31 23 567 9699
NL-2132 NG Hoofddorp               Main Office: +31 23 567 9600
---------------------------------------------------------------