On 3/22/07, Venkata Krishnan <[EMAIL PROTECTED]> wrote:
Hi Jim, Thanks for sharing this information - its really useful. - Venkat On 3/22/07, Jim Marino <[EMAIL PROTECTED]> wrote: > > Hi, > > We just finished the ServerSide demo and I figured I send a mail to > the list outlining how it went... > > We had the slot following the opening keynote and were up against Rod > (Spring) and Patrick (OpenJPA) as the other two talks. I was > surprised to find that the ballroom was pretty full. I gave the talk > and the demo showing end-to-end federated deployment and reaction > seemed very positive. Meeraj gets the "hero" award for staying up to > an obscene hour in the morning to implement a JMS-based discovery > service as we encountered last-minute hiccups with JXTA. > > My observations are: > > - After speaking with people after the presentation, feedback on the > value of SCA was consistent. Specifically, they thought the > programming model was nice but not a differentiator. What people got > excited about was being able to dynamically provision services to > remote nodes and have a representation of their service network. In > this respect, I think the demo worked well. Two people said they need > what the demo showed for projects they currently have underway. > > - People asked how SCA is different than Spring. They reacted > positively when I said "federation" and "distributed wiring". Related > to this, people get dependency injection (i.e. it's old-hat) and just > seem to assume that is the way local components obtain references. > > - People seemed to react positively when I compared SCA to Microsoft WCF > > - People liked the idea of heterogeneous service networks and support > for components written in different languages, particularly C++. > > - People didn't ask about web services. People were nodding their > heads (in agreement) when I talked about having the runtime select > alternative bindings such as AMQP and JMS. > > - People want modularity and choice. Two areas they wanted choice in > was databinding and persistence. They liked the fact that we are not > locked into one databinding solution and that we have JPA > integration. (as an aside, they also liked that SDO can be used > without SCA). Spring integration was also popular. > > - People also liked the idea of a 2MB kernel download. One person > mentioned they only want to download what they intend to use and not > a lot of extra "clutter". > > - People wanted to know how SCA is different than an ESB. I basically > described it using the switch vs. router metaphor and how a component > implementation type can be a proxy for an ESB. Related to this and > point-to-point wires, people thought wire optimization by the > Controller was cool. > > - People seemed to be more interested in running Tuscany as a > standalone edge server or embedded in an OSGi container. I didn't get > any questions about running Tuscany in a Servlet container or J2EE > application server. This seems to be consistent with there being a > number of talks on server-side OSGi. > > My big takeway is that we need to make the demo a reality. > > Jim > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > >
Jim, Nice one. Thanks for the summary. Did the conference record the talk? Would be good to see it. Noting your comment and recent mails about the last minute changes to get JMS working in short order, is everything checked in that's needed to run the demo? Looking back I see several notes on build instructions and it would be pretty cool to give it a spin. Can I ask a question about support for components written in different languages? Did people specifically say they were interested in C++? Did they mention other languages (and, if so, which ones)? Presumably the sweet spot is the ability to show components implemented in various languages all acting as part of a single SCA Domain. How big a deal do you think this ability to be able to draw a "picture" of you heterogeneous service network (in SCDL) vs some of the other things you mention like "standalone edge server" or "selectable bindings". I'm asking this question because, as you know, I like the idea and from your notes it seems the audience likes the idea but I'm interested to know how much interest there was for this vs other things. I imagine, from reading your closing comments, you have a whole stack of ideas now in your head about what needs doing next. This would seem like a great opportunity for us all to look at what technical challenges lie ahead and to have a discussion about how, as a community, we step up to meeting some of them. How do we do this? Do we start some threads on individual items? A thread on the grand plan and then split onto areas of peoples interest. Having this summary is great because is really pushes on what we really need to focus on, i.e. making something that is useful to our (potential) users. We need to convert it into technical opportunities that get the creative juices flowing. Regards Simon