Re: [xri] Comments on draft-resolution-04

[Peter C Davis <peter.davis@neustar.biz>]

On Thursday 03 March 2005 01:00 pm, Wachob, Gabe wrote:
> Peter-
>  Some questions on your feedback. I couldn't figure out your line
> # references - they seem different than the PDF version of -04. But I
> was mostly able to track. One comment I didn't track (ie couldn't find
> the text you were referring to):
>

hmm. i was using the xri20res.pdf i located in kavi...glad you found them 
nonetheless.

> "line 1025: for 'reasonable' caching, whatever that means, it is
> unlikely that
> authorities can/should be aware of down-stream cache duration
> directives, thus
> i think the advice concerning the cache relative to the resolution chain
> should be dropped, and we should simply point out that cache times may
> vary,
> and the resolver should simply adhere to the 'shortest-lived' cache for
> the
> resolution sequence."
>

seems that it should have been closer to 1015 (int he above ref'd doc).  but 
what i was trying to point out is i found little advice for lookahead 
resolver deployments on how best to compute the HTTP cache header based on 
the each of the XRID's it got back.  so, follow this pseudo conversation:

client: resolve =a*b*c
lookahead server:   GET /resolve/*a
A-Authority: 200 OK Expires March 01 <Expires>March 01</Expires>
lookahead server:   GET /resolve/*b
B-Authority: 200 OK Expires March 02 <Expires>March 02</Expires>
lookahead server:   GET /resolve/*c
B-Authority: 200 OK Expires March 03 <Expires>March 03</Expires>

lookahead server to client: 200 OK Expires March 01 
<XRIDescriptors>...</XRIDescriptors>

so the lookahead resolver (server) MUST (i think) use the nearest expiration 
header for it's expiration response to the client.

make more sense?

--- peterd