Per-Entity Metadata Working Group

["Cantor, Scott" <>]

On 9/14/16, 4:35 PM, "Nick Roy" 
<>
 wrote:

>    I thought I remembered you saying that you were going to have to make 
>    changes to Shibboleth client software to achieve the kinds of uptime 
>    goals you were hoping for when we started researching CDN 
>    characteristics.  What happened to that line of thinking?

That line of thinking was really foremost about caching across restarts, and 
yes, we will (Brent just sent his IdP proposal to our list).

But I don't really see us implementing the levels of sophistication in the 
HTTP client code we'd be talking about here.

Failover is "free" as long as you can live with timeouts, but I think that's 
somewhat unreliable in practice (timeouts in networking code are very hard to 
trust, at least in my experience).

But I don't know that we'll be implementing redundant URLs inside a single 
resolver, and I definitely don't see us implementing multiple A record 
support on a single hostname unless libcurl and/or the Apache Commons 
HttpClient already do it for us (they might, I haven't looked).

-- Scott