Oisín Hurley's Weblog

old dog, new tricks

Archive for April 2009

Intermittent Bug in Eclipse Community

leave a comment »

This bug has been cropping up on and off since Eclipse 3.2. The usual pattern is that it becomes visible early in the year to the Planning Council, who immediately attempt to triage it and limit the potential solution space so that resolution can be made as cheaply as possible. Because Eclipse is an open organization, and is populated by humans, who are in the main scurrilous gossipers and rumour-mongers, awareness of the bug expands in the community as a whole. A Bugzilla is usually created to give the community the opportunity to cosmologically inflate the solution space through the time-honoured approach of getting your oar in.

Eventually, broad interest wanes, and the determination comes back the Planning Council again, who, disappointed that no really clear solution has come out of the community involvement, many eyes == many opinions it appears, just sigh and put a workaround in. Next year the same sequence of events occur.

This year, it’s bug 271054, but the problem is the same.

What do we call the next roll-up release after Galileo?

Comments about bike sheds will be modded up appropriately.

Written by oisinhurley

April 3, 2009 at 2:40 pm

Eclipse 2009 Thursday

with 5 comments

Nearly there. This is the final article in a tetralogy of tardy tales from EclipseCon 2009. I don’t know about you, but I’m looking forward to the end of this, where I can put a big scratch mark through my notes and call them done.

Thursday 26th March

OSGi Dual Talks: Clouds + Bundle Generation

This dual talk turned into a singleton due to the US immigration office who didn’t manage to get a travel visa to the presenter in time for the conference. There seemed to be a lot of this about. Because I’m on the programme committee, I got to see a number of emails, and there was a couple of tweets too, where developers with Eastern European-style names didn’t get replies on their visas in a timely fashion. Further research indicates that the immigration people are guiding 150 days in advance application for scientific/technical conference travel visas. So, as the Irish Girl Guides say bí ullamh – be prepared: either get your application in early, or change your name.

Back to the cloud. Alexsey showcased some interesting tech around managing and deploying bundles into Amazon EC2, as well as discussing some of the security issues. There’s a lot of this about too. Many companies are constructing their own tools around deployment and management into Amazon and private clouds, using OSGi technology as a handy enabler for standardized packaging and lifecycle. The scale of the activity was brought into sharp relief when I got to sit in on an OSGi Tools meeting on Friday morning. Conclusion – moderately interesting. Next step – wait for someone to develop a standard application model for OSGi.

Conquering GEF: Creating well designed graphical editors and bringing them to the Web

Over in FUSE-land, we’ve got a lot of software that use graph models. Integration graphs, dependency graphs, pipelines, that kind of thing. Graphs are great for visualizing, but when you’ve got the job of doing this for both browser and IDE based tools, then you start looking for ways to minimize your code base forkage. The RAP guys have done a great job in bringing SWT to the Web. If there was a chance that GEF, or better still Zest, could be brought to the Web, then that would be full of win. That’s why I went to this talk, which was standing room only.

Once more, this was a talk of two halves. The first half was a kind of a playbook for GEF, and while that was moderately interesting from the point of view of a useful summary of the trials and errors of getting GEF right, the real meat was in the second part of the talk. Here, Vineet brought us from a GEF editor running in a workbench, to the same editor working in a standalone form without a workbench, to the same editor working in a browser. Very cool. The only downside here is that it’s early days for the work, and there are issues that need to be resolved as well as features that need to be added. The way that it happens behind the scenes is that the original GEF code is compiled to ActionScript (some manual intervention required if there’s name clashes) and then that ActionScript is the thing that gets rendered in the browser window. Definitely worth looking into as it matures. Conclusion – very interesting. Next step – come back and check this out after the summer.

Eclipse Swordfish – an open source SOA runtime framework for the enterprise

Swordfish is Eclipse’s very own ESB – you can’t have an open source community without one – and builds upon Apache ServiceMix and Apache CXF with some added-value capabilities and services. One neat technological advance they have in there is an extensible framework for dynamically generating interceptor chains which are installed in the Normalized Message Router to mediate message exchange. Now there’s a statement that only makes sense to probably about ten people, all of whom are already too busy writing in their own blogs to notice it and marvel at its awesome niche techiness. So it goes. For everyone else, I’ll offer the statement that this enables policy-driven behaviour. The ESB is OSGi-based, and uses JBI (also not as dead as people think) as the way to bundle up units of function. Conclusion – very interesting. Next step – must try it out a bit.

Afterwards, I said to Oliver that it was a good presentation, but probably a bit too much technical detail towards the end. He reminded my that the last time he did a presentation I had told him it was too high-level. He then asked was it possible to make me happy. I replied in the negative.

There’s always an issue with demonstrating middleware. It is not interesting to some people to see command windows and logs – the MEGO effect is immediately apparent. It’s like demonstrating intestines. All you have to look at is the output, and it’s not really that pleasant. Ideally, you could take a chunk of intestine out, and attach electrodes to it, then you could pop your microphone in one end, throw the power switch and the induced peristalsis would rocket the mic out the other end, where you could catch it in an amusing manner. I bet you would get a whole lot of +1s for that demo. But you can’t do that with middleware, alas.

SOA Ecosystems

Yet another one of these curated talks – I’ve got feedback on the plus, minus and meh sides of the argument on this model for talks, so I’m interested in what y’all think. Leave a comment with your, er, comment. I’ll bring it back to the programme committee.

This started off badly because I turned up late to introduce it. Zsolt had already started Eclipse Community Registry, a building block to foster the adoption of the Eclipse Runtime and I’d swear that I got a few filthy looks from Adrian. Sorry, guys. This community registry that Zsolt was talking about sounded like a super-duper version of EPIC – but for services. It’s a service registry with handy stuff like tagging, voting, commentary and the like. Zsolt made sure to point out that this is just an idea at the moment, it’s not a project as yet, he’s just putting it out there to see what people think. Conclusion interesting. Next step – wait and see.

The second talk, galaxy, an open agile platform using dynamic software architecture continued the theme of long talk names, and removed the convention of capitalization. This was a consumer style of talk – Fy brought us through a fairly large-scale initiative being explored by INRIA to construct a development platform that integrates open standard technologies, includes an agile design and modeling environment, and ensure direct feedback to the design from a monitoring infrastructure. It looks neat, and my first thought was that well, this is a long way off. But it turns out there’s a date on it, in Q4 2010, so that makes it more interesting. Conclusion – very interesting. Next step – keep an eye on what’s going on.

Personal applause for Fy because he did a presentation which was almost completely pictures! Yay!

The final event of the day, and the conference, was The Eclipse Community Spotlight. It’s the same every year, the top-level project leads, or their designated PMC representative get to sit in front of the whole crowd and answer questions. Every year, Doug Gaff does the lyrics of Baby Got Back sotto voce into one of the mics before they are switched on. This year, we found out exactly how much Eclipse real estate David Williams (WTP lead) owns – all of our base is pretty much belong to him.

That’s it for this year. As per usual it’s all great fun, and there was visits to one of the best Thai restaurants in the area, one or two great socializing occasions at receptions and bars, and awesome high-bandwidth interactions with people that can help me solve issues, forge alliances, explain concepts, and such important things. Ultimately, huge thanks are due to Bjorn, Scott, Anne and the rest of the extended team that made this all happen.

(Don’t forget to take the survey!)

Written by oisinhurley

April 3, 2009 at 1:05 pm

Posted in conferences, eclipse

Tagged with

EclipseCon 2009 – Wednesday

with 2 comments

Today I continue my week-in-arrears retrospective of the EclipseCon 2009. Authors often mumble into their whiskey about the tyranny of the blank page, but I feel that the full page (of partially illegible notes) is equally tyrannical. And I have no whiskey either.

Just a quick concerned citizen-o-gram by the way, please go and fill out the EclipseCon Survey, paying special attention to the the areas that you would like to see improve. It’s important for the program committee next year.

Wednesday 25 March

Building Applications for the Cloud with Amazon

This was the keynote first thing in the morning. I was impressed with Don MacAskill’s shoes, and the way that they run SmugMug on Amazons cloudtastical offering. I experience a veritable frisson at the new EC2 deploy and debug tools that Amazon have created for Eclipse. It uses Webtools’ server framework and deployment mechanism to allow you to deploy your webservices to machine instances running in EC2, and it allows you to debug them as they run. Very neat. Limited to Tomcat at the moment, I’m sure it’ll be no time at all when you will suddenly see it working for things like FUSE ESB and FUSE MR to name but two, ’nuff said. Conclusion – interesting plus frisson. Next step – extend Amazon code to deploy to FUSE ESB instances.

What’s Baking in the WTP Incubator

was a little curated number intended to act as a showcase for some new projects that are growing in the WTP Incubator project. In a session of two halves, Shane Clarke gave us an insight into Developing JAX-WS Web Services – this is a new piece of tooling that enhances the dynamic web project approach of WTP so that it’s easy to create JAXWS services. I’m biased, since we consume this stuff in FUSE and I’m a committer on the project, but Shane has done a maximum awesome job here, I can’t recommend the man highly enough. Unfortunately, this looked like it was a two-hour talk that had to be jammed into twenty-five minutes, so I’m sure that there’s plenty we didn’t get to see. Conclusion – awesome. Next step – update the Apache CXF wiki with details of this.

Next element in the showcase Incubating XML Security Tools. This incubator project contains a set of capabilities for dealing with various XML Security standards, like partial encryption, signing and the like. The author, Dominik, originally constructed this as an e-learning thing, for people that wanted to experiment with XML Security and see how it all worked. It works pretty well for that, but I’m not sure if there are public APIs in place which would allow you drive the encryption processes programmatically. That would be of more value than the user-driven version. Conclusion – interesting. Next step – somebody translate the help from German please!

From source to automated builds with Buckminster and p2

Buckminster is the software assembly tool that we use to do the STP builds, and it has served us well over the past couple of years, so I was interested to see what has happened recently, given all of the p2 stuff going on, and Eclipse release train requirements around packing and signing and all that. Good news – there’s a recent version of Buckminster that is p2ized and fully kitted out for packing and signing headless builds. Just what the doctor ordered. I would rush off to try it out immediately, but I’m stuck here writing in this damn blog. A footnote on the presentation – the subject matter is a little esoteric, but Henrik did a good job explaining it in a way that didn’t dive deep into the details – except perhaps with the CSPEX thing. Conclusion – w00t. Next step – update to the Magic Buckminster and plumb in my STP generic build system, then blog about it.

Higher-level UI programming

Intriguing title on this session. Perhaps we were to be shown how to program UIs by pure radiant thought alone? Or perhaps we were going to see some Epic-class characters showing us how to do our UI? This was a two-session talk again. First up was UFaceKit – A highlevel Databinding and Widget-Toolkit-Abstraction – the motivation behind this particular piece of work is to provide a higher-level widget toolkit abstraction that can be successfully mapped to alternate widget toolkit implementations, like Swing, GWT, SWT and the like. UFaceKit is now an Eclipse project. Conclusion – interesting. Next step – keep an eye for an opportunity to experiment.

The second talk in the showcase was UI designers : Untangle the knots, use EMF (live) models! (yes, there was an entertaining selection of session titles this year for some reason). We got to see an EMF model representing an SWT UI rendered, live, with instant updates to the UI as you tweaked the model. Very nice. The code that does this is an open source declarative UI framework (ye gods, another one) called Wazaabi. Conclusion – me like! Next step – use this for kick-starting mockups at least.

Eclipse and Maven

This next session was one that I was particularly looking forward to. Eclipse, Maven and how they interact is a persistent thorn in my side – always has been – so I was going to this session in the hope that there will be some release from the anguish. There are two Eclipse and Maven projects at the Foundation, and this session was split into two parts to align with the projects. Some people came along hoping for celebrity deathmatch – I just wanted some solutions.

I’m not going to go into the details of the sessions – I can summarize with m2eclipse FTW, apologies to Phil and Brett and Carlos, but m2eclipse is really ploughing resources into this and is moving ahead with a much more complete and broad solution. Conclusion – m2eclipse FTW. Next step – I think I’ll have to start committing to this when it moves onto Foundation turf.

I got the chance to have a sit-down chat with Jason of the m2eclipse project later in the day and we talked about where it was going, and I asked some questions about how I can incorporate the thing into the FUSE Tools and meet the double goal of providing a means for developers to move in a frictionless manner from snappy Eclipse builders to Maven CBI approaches and not suffer terrible brain-damage. Jason gave me enough confidence that this will happen, but it’s going to require some work. I’ll pick up that thread again in another blog entry.

Written by oisinhurley

April 2, 2009 at 10:51 am

Posted in conferences, eclipse

Tagged with