Re: Clarification on Alternative Service Connection Failures

On Fri, Sep 29, 2017 at 3:50 PM, Ryan Hamilton <rch@google.com> wrote:

>
> ​One more thought. Should non-idempotent requests be resent in the case
> where an alternative 500s?​
>
>
The answer is the same whether or not there is an alternate.  The
alternates are semantically equivalent to each other.

You can say that its inconsistent to say that and also that a 500 might
make you flush the alt-svc entry from the cache, but you can flush the
entry any time you don't like it (too slow for example).