This sounds reasonable to me, although I think that we probably want
to use the prior knowledge approach for the majority of tests, rather
than using upgrade (unless upgrade is being explicitly tested).
Stuart
On Thu, Apr 20, 2017 at 7:40 AM, Edward Burns <edward.burns_at_oracle.com> wrote:
> Hello Volunteers,
>
> This email only applies to EG members that plan to run the Servlet 4.0
> TCK.
>
> Unfortunately, in spite of our best efforts, there will be no JDK 8
> backport of ALPN. The resolution is the ALPN implementation is left as
> an implementation detail of the Servlet container. If we ever meet out
> there on the conference trail, I'm happy to talk about it over a beer.
>
> But, in any case, this leaves a problem for the TCK. The TCK must be
> runnable under any Java SE 8 compatible JVM. Traditionally this applies
> both to the TCK client and to the JVM running the Servlet Container.
> The latter case is covered by the above resolution. The former is still
> an open question.
>
> This email proposes we modify the TCK such that it only exercises the
> HTTP/2 features over h2c. In this way the ALPN resolution can stay
> within the servlet container and not pollute the TCK client.
>
> This appears to be the least bad option given the unfortunate resolution
> of the ALPN backport request.
>
> ACTION: Please reply by start of business Tuesday 25 April 2017. If
> there are no responses, we'll go with this proposal.
>
>
>
> Thanks,
>
> Ed
>
>
> --
> | edward.burns_at_oracle.com | office: +1 407 458 0017
> | 1 business days until planned start of Servlet 4.0 Public Review