[wildfly-dev] ALPN support

[Brian Stansberry]

On 1/11/15 8:00 PM, Stuart Douglas wrote:
> So I guess the real issue for domain mode is how we actually tell the
> configured servers to use the ALPN jar, especially given that this
> situation will likely be temporary, and not be required for Java 9+.
>
> One option could be that if the -alpn option is passed to the host
> controller on boot then all servers that it spawns will also have alpn
> on the boot class path, but that just seems really hacky.
>

The launcher stuff that James has done is meant to allow mirroring of 
our script behavior via a java API. So if we're adding stuff like this 
into the scripts it should go into the launcher too. Otherwise tools 
can't replicate script behavior. At least not without coding it up 
themselves.

If it's in the launcher API it's not that hacky to pass the flag through 
to the HC process so it knows to use it.

The whole thing is quite yuck though so (no offense meant) so I feel 
squeamish advocating putting it in the launcher API.

> Another option would be to just require the user to setup the correct
> -Xbootclasspath option manually in the JVM arguments. This is not very
> user friendly though.
>

A third yuck option is to use RuntimeMXBean.getBootClassPath() and do 
hackery to identify "foreign" stuff, and then have the PC use that when 
creating the HC process and have the HC use it when creating server 
processes. But I like just having people configure it better. It's a 
pretty edgy case and it falls within the parameters of how a jvm config 
is meant to be used.

I think the bigger question is whether this ends up in the launcher API.

> Stuart
>
> On Wed Jan 07 2015 at 2:39:27 PM Jason T. Greene
> <jason.greene at redhat.com <mailto:jason.greene at redhat.com>> wrote:
>
>     We should probably download these jars using a directory or file
>     name convention that is associated with the particular jvm version.
>     That way it becomes easy to support multiple jvm version
>     environments which is very common (both in domain and standalone).
>
>     In domain mode we could have the host controller bootstrap follow
>     the same approach as standalone. For domain server configured jvms
>     we could reuse the logic as part of the JVM argument building process.
>
>     On Jan 6, 2015, at 7:20 PM, Stuart Douglas
>     <stuart.w.douglas at gmail.com <mailto:stuart.w.douglas at gmail.com>> wrote:
>
>>     Hi all,
>>     Both SPDY and the upcoming HTTP2 protocol require the use of an
>>     SSL extension called ALPN (Application Layer Protocol
>>     Negotiation). Unfortunately this is not supported in current JDK's
>>     (it should appear in JDK9), so the only way to support these
>>     protocols in Java at the moment is to modify the boot class path
>>     and override the JDK SSL classes to add support for this.
>>
>>     In practice this means add jetty-alpn-boot.jar to the boot class
>>     path, however the exact version of the jar that is required
>>     depends on the JVM version, which means we can't just ship a jar
>>     and add the boot class path stuff to our startup scripts.
>>
>>     I have come up with a proof on concept[1] for how to deal with
>>     this, that will download the appropriate ALPN jar for the current
>>     server if -alpn is passed to the startup script, however we still
>>     need some way to handle this in domain mode, where we need to make
>>     sure the servers are all started with the appropriate parameter
>>     (worst case we could just require users to install the appropriate
>>     JAR themselves, and then set the appropriate JAVA_OPTS).
>>
>>     Hopefully the need for this will go away with Java 9, so I am not
>>     sure how much effort we should spend dealing with this.
>>
>>     Does anyone have any thoughts on this?
>>
>>     Stuart
>>
>>     [1]
>>     https://github.com/stuartwdouglas/wildfly-core/compare/wildfly:master...stuartwdouglas:alpn
>>     _______________________________________________
>>     wildfly-dev mailing list
>>     wildfly-dev at lists.jboss.org <mailto:wildfly-dev at lists.jboss.org>
>>     https://lists.jboss.org/mailman/listinfo/wildfly-dev
>
>
>
> _______________________________________________
> wildfly-dev mailing list
> wildfly-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/wildfly-dev
>


-- 
Brian Stansberry
Senior Principal Software Engineer
JBoss by Red Hat