Getting low level implementation detail focussed for a second I'd say we could do all of these things...
i) make all the ServiceMix goodness available as bundles/jars that folks can use on any classpath or OSGi container (pretty much what we do already really - though we maybe need a couple of tweaks to work easier outside of OSGi maybe - or PojoSR maybe - or maybe something a little lighter when outside of OSGi?) ii) provide a Karaf distro with all the ServiceMix goodness in there; preferably light weight with some stuff optional features to keep things small & light (e.g. making NMR/JBI an optional feature etc) - (again kinda what we do already just maybe doing a smaller lighter version for 5) iii) provide a WAR using ServiceMix with some permutation of ActiveMQ / Camel / CXF / ... etc - or make it easy, say via maven WAR overlays, for folks to bind ServiceMix goodness into their WARs iv) provide a Tomcat distro with ServiceMix inside - a ServiCat / TomMix thingy; where servicemix goodness is available on the shared class loader (with maybe a separate WAR for the web console or something) so that if any WAR is deployed with some ActiveMQ / Camel / CXF / ... etc stuff inside, the ServiceMix stuff can do its thing; without every WAR having to include a full ServiceMix distro inside it. Option iii) is pretty easy and would work in any Servlet engine today - assuming we do (i) so that the ServiceMix goodness can work on a simple flat class loader (e.g. PojoSR today - or maybe something a little lighter if OSGi isn't a requirement); the downside is it means including all the same stuff in each WAR. Whereas (iv) would be specific to the ServiceMix Tomcat distro (possibly based on TomEE?) but would allow hopefully any WAR using ActiveMQ / Camel / CXF / ... to just work with the ServiceMix goodness. For folks using multiple WARs and ServiceMix goodness option (iv) would be the lightest option, at the cost of a custom Tomcat distro. Incidentally maybe one day the TomEE extension mechanisms could become part of the standard Tomcat distro; then its really we'd have 'Tomcat extensions' for EE / ServiceMix; then folks could take a vanilla Tomcat distro and use the Tomcat extensions mechanism to install EE (OpenEJB) or ServiceMix? (Without grokking TomEE the previous sentence probably wont' make much sense but don't worry about that, it might take a long time to get there anyway...). On 1 July 2011 21:01, Jean-Baptiste Onofré <[email protected]> wrote: > Hi all, > > I'm not agree with that, and I don't think that it's the Guillaume's point > of view. > > The idea speaking about Tomcat is not to create ServiceCat or something like > that. If the purpose was this one, Karaf + Jetty already does the stuff. > > The idea is more to provide: > - ServiceMix as a standalone ESB (as ServiceMix 4) powered by Karaf > container > - ServiceMix as an embeddable WAR or EAR. Tomcat is just an example but the > purpose is wider. For my point of view, if we want to go this way, we > shouldn't be focus on Tomcat but be more generic. The purpose is to be able > to deploy ServiceMix 5 into existing containers like Tomcat, but also > Glassfish, Websphere, etc. The goals are to use the container resources > (JNDI, JTA, etc integration) and increase the ServiceMix adoption for > existing IT administrators. > > Regards > JB > > On 07/01/2011 09:03 PM, Michael Van wrote: >> >> Upon re-reading this thread, it appears there are two subtely different >> approaches being suggested: >> 1) To merge most of SMX into the Tomcat container creating a new little >> beasty "Servicecat" or something, and >> 2) To modify the existing SMX bundles so that they work inside of a (any) >> servlet container. >> >> I like the second option more than the first, because it will greatly >> increase the set of users that can/will-be-able-to use it. The first >> option >> is intriguing, but I dont' know if it would overcome the stated design >> goal >> of allowing folks who are using legacy environements to use SMX. >> >> Could someone clarify the option being proposed? >> >> >> dblevins wrote: >>> >>> >>> Michael Van wrote: >>>> >>>> From a geeking out perspective, the daunting task of moving SMX into >>>> Tomcat seems like a good challenge to take on! Overcoming the >>>> traditional issues of war-war communication using RMI would be tough, >>>> but >>>> the result could be a better way of doing things inside of servlet >>>> containers, which I can see being adopted by a very large community of >>>> developers. From that point of view, I can see that this move will >>>> provide quite a bit of new utility to all servlet developers. That >>>> improvement, along with the ability to run a fully-fledged ESB inside of >>>> a servlet container would add a significant tool to a servlet developers >>>> toolbox. >>>> >>> >>> Maybe two years ago Tuscany got inspired to do the same with Tomcat and >>> also pointed at what we now call TomEE. They basically took all the >>> Tomcat integration code, said thank you, and went their separate way with >>> it. I admit it was a little disappointing as we would have happily >>> supported them and it would have been great to see the idea grow rather >>> than be copied. It's been forked again recently for a commercial app >>> server. Similar thanks and see you later. >>> >>> We're more than happy to support ServiceMix if you guys want to go down >>> this route. The backup plan could of course be to take the code and run, >>> but it would be great to at least give working together a shot. >>> >>> >>> -David >>> >> >> >> -- >> View this message in context: >> http://servicemix.396122.n5.nabble.com/DISCUSS-Rebooting-ServiceMix-5-tp4528896p4543013.html >> Sent from the ServiceMix - Dev mailing list archive at Nabble.com. > -- James ------- FuseSource Email: [email protected] Web: http://fusesource.com Twitter: jstrachan, fusenews Blog: http://macstrac.blogspot.com/ Open Source Integration and Messaging
