I've been talking about the ideas behind JBossEverywhere for a long time and diving into details during and after the keynote. I've also presented on these concepts since to several JBUGs and other groups. Most of the people I've spoken with just "get it", but there have been occasions where I've had to dive into a bit more detail to fully explain what I mean: some things that may seem obvious to me aren't necessary that obvious. Plus I've been asked that despite the vision, what is our stand or position? Well hopefully this article will settle things so we can move on and get this done. Before I go on though, it's worth stating that this will be an unapologetically Java and JBoss specific article: I can make many of the same arguments you're about to hear more generally about middleware, but I'm not going to do that!

OK what is our position? Well let's look at some positions or visions that have been attributed to others in the same area we operate. One of them is that Cloud is the death of middleware. Hmmm. Now if that statement had been made by vendors who don't have a vested interest in persuading you to move away from existing middleware implementations to their own, or who don't have any investment in middleware, I might give it more than a few seconds thought. Now of course you could say that given my position, I'm hardly being objective either, and to a degree I can see your point. However, I've lived through 3 decades of middleware standards and implementations, including DCE, CORBA and J(2)EE. My background is heavily influenced by scientific method and analysis, so I try to be objective even when it may go against perceived business sense.

So let's just assume that for the rest of this article I'm writing with my professorial hat on and not my Red Hat, OK? And let's think about that statement: "Cloud is the death of middleware". What does it really mean? I think we first need to look at what is meant by middleware, which is pretty poorly defined. Given the vendors involved, I suspect it's more Enterprise Middleware that is assumed. So what is being suggested is that security, transactions, reliable messaging, availability etc. are no longer required in the Cloud? Give me a break! Let's look at a human-centric equivalent of (public) Cloud: outsourcing. If you've ever been involved in outsourcing something, such as IT or a call centre, then you'll have drawn up a set of questions or rules by which the prospective recipient who will run the outsourced effort must be held. These would include looking at their plans for high availability, security of your data (particularly if they host data on behalf of your competitors) and especially how to get in contact with then 24x7 in the case of a problem. If you don't have such a set of criteria and you have outsourced then please let me know so I can make sure I never use you and that if you have any of my data then I can scrub it from your system now!

Public Cloud, or at least public PaaS, needs good middleware and needs it now. When you look at what PaaS really is, then defining it as Middleware as a Service is not going to be far off the mark. Why isn't it defined this way? Well just look at where the term PaaS comes from and that will give you your answer. Middleware techniques and protocols have been developed and honed for well over 4 decades and in many cases haven't changed for years. Of course Cloud offers new problems and challenges, such as multitenancy, but some things remain fundamental. And what about scalability? Hmmm, have you ever looked at the Web? That's a pretty big system! It dwarfs any Cloud data centre and works pretty well. So let's put this one to bed: Cloud isn't the death of middleware! Far from it.

Before moving on, let's tackle a related statement: that Java Application Servers are either heavyweight or dead. Again, this isn't being said by people in an objective manner. I've already discussed this issue separately, and you should take a look at that for completeness. However, the general gist of that article is that a good implementation of the Java EE specifications need not be heavyweight, bloated or irrelevant to applications that don't need all of the enterprise capabilities. Java EE 6 and beyond are the right stacks to consider, both for small scale and large scale applications, and specifically JBossAS 7. Maturity of implementation is critical in these areas for very obvious reasons, and we have a set of best-of-breed components that other vendors are trying to replicate and continuing to fall short on their aims! The concept of the Application Server will evolve over time, pushed by requirements and restrictions on different deployment environments, but let's not try to kid ourselves that somehow it simply isn't needed: it is, but what *it* is tomorrow will look very different from what it looks like today. And guess what? JBossAS 7 is *it*!

Another position I've heard recently is that Java needs to make itself relevant in the Cloud or cede the space to languages such as Ruby. That's so profound and visionary. (Where did I put those sarcasm markups?) That's about as obvious a statement to make as, say, a football coach telling his team that "If we score more points than our opponents then we'll win." Come on, surely we can do better? If any language cannot adapt to changes in the industry then it will be replaced by those that can. That's the way it's been for decades, and no I'm not going to drag up the COBOL reference (oops!) Java, the JVM and Java middleware is already relevant to a large part of the community that wants to move to the Cloud. Of course there are challenges, and I mentioned some of them before, such as multitenancy, which need to be better addressed in the language itself. Statements like that may make good soundbites, but not much more. (I really wish I could remember where I mislaid my sarcasm markups!)

As a community I have faith that we'll make those changes and we are definitely involved in pushing those changes forward at a much quicker pace than we've been used to before. But the Cloud, and specifically PaaS, isn't going to be a mono-language area. There will be many different languages for different use cases. But where we stand on this should be fairly obvious, particularly if you roll up both of the positions we've considered so far: for Enterprise PaaS you need Enterprise Middleware, and if CORBA taught us anything it's that you can have a single runtime written in a language that isn't used by your developers and as long as that runtime is good (efficient, reliable, manageable etc.) then your users/developers won't care. They'll see all of the advantages and none of the disadvantages that they associate with the underlying language. We're seeing that already with projects such as TorqueBox, where Ruby developers can use the application server without ever having to know Java. And we'll be seeing more of these kinds of projects from JBoss in the future, but read on.

But even looking at these previous two statements they are far too narrowly focused. Yes Cloud is important today, but it's not where we should be putting all of our attention. That needs to be in the area of ubiquitous computing, which should subsume Cloud and mobile. Let's look at mobile for a few moments, and with apologies to others, I'll state what seems to me to be a fairly obvious fact now: there will only be two dominant platforms in the future; iOS and Android. Of course there'll be needs for niche players, but the old order has fallen. Smart phones will dominate and the types of applications that people want to write on devices, such as phones or pads, are already increasing in complexity. Native applications aren't going away any time soon. Yes, I know this kind of bucks the trend, with others pushing for thin clients in general such as the Chrome Book (and yes, I know this isn't for a mobile phone but stick with me on this). But look at it this way: if all we wanted to do was run a browser on the phone to access applications or data that are hosted elsewhere, then we probably should have stopped innovating with the hardware a few years back!

There are billions of phones on the planet. But there are tens of billions of processors, including those in your washing machine, games console or car engine management system. Many of them are more powerful than workstations from 2 decades ago! This is ubiquitous computing. It's pretty obvious that the types of applications that will eventually be written for them will grow in complexity and store more information locally. (I've lost count of the number of times I lose wireless or phone signal on my phones and pad!) Application developers, whether on the phone or embedded devices, will want the kinds of capabilities that we take for granted in the JBoss world, including messaging, transactions and security. This is ubiquitous enterprise computing!

So now on to the second part of this stand ...

I mentioned earlier about presenting at a couple of JBUGs this week. Fortunately both presentations were on the same subject: where next for JBoss. Both events were packed. I'm not sure of the exact make-up of the audience in London (and thanks to Atos for hosting!), but in Newcastle we had people from the University as well as several local companies, all either existing users of our projects or considering using them. In Newcastle we managed to duplicate the keynote demonstration too, though this time without the mobile device UIs that we showed in Boston. It went very smoothly! There were some good questions too around some of the underlying aims behind JBossEverywhere and hopefully I persuaded some of the new attendees to come back to future events.

At the London event I didn't have the luxury of the hardware we used for the demo, so instead I took a copy of the official video. Again it was a packed room, with a number of JBoss/Red Hat people there including Manik Surtani (who was wearing almost exactly the same outfit as he did in Boston!) So once I ran through the slides, we watched the video and I took questions again. Some of the key concepts seemed to resonate with folks, although there were specific questions around Cloud and how it relates: as I tried to say in the video, I think this is the future of Cloud, so our current strategy with efforts such as OpenShift are complimentary. Plus as I pointed out during the talk, it's good to have a vision to catalyse people and focus their efforts.

I enjoyed presenting at both events. It was good to hear comments and suggestions. I know that this theme is something that I'll be focussing on quite a bit for the rest of this year, with several presentations already submitted to a variety of conferences and workshops. And maybe other JBUGs if there's interest!

If you've watched the video of my JBossWorld keynote or followed other presentations I've given around the future of JBoss or middleware in general, then you may have heard me talk about JBoss technologies evolving into a fabric, or ubiquitous middleware that is always there, in the background, no matter what device you may be deploying your application on. The name "fabric" springs to mind, though it is used by many others to mean different things. I suppose another word to use might be "ether", given the historical use of the word in the physics community. (OK we would need to ignore the fact that the belief in the ether ended when Michelson-Morley's experiment proved it did not exist.)

Let's stick with Fabric for now though. Well at the Red Hat Partner Summit I was giving a presentation on The Future of JBoss and was talking about JBoss as a Fabric. I've used a number of analogies to try to illustrate what I mean by this, including the distributed operating system one I've used before. Another one I used was that of TCP or UDP: capabilities that we all just take for granted these days, no matter where you're deploying your application these days. Now of course not all of the JBoss capabilities (projects) will be available in all environments. So in the grand tradition of cloud (to which we are most certainly not tying Fabric), I suppose we are talking about Just enough Middleware (JeM), Just enough JBoss (JeJ), or Just enough Application Server (JeAS) - though I've used that one before.

So what next? Well I'm fairly sure I want to create a new project for Fabric so that the wider community can get involved, though I may wait until after AS7 goes final. That means a project name of course, which is a problem as I'm not so good with names. Maybe Aether?

Unless you've had your head in the sand for the past 18 months then you can't fail to have noticed the releases of JBossAS 6 and then the move through various betas of AS 7. Well we are on the final leg of what at times has seemed like a marathon and at others like a sprint. The team, which includes the various individual project teams, productisation teams, docs and QE, have been working flat out and I can tell you all that once we are finished with the EAP 6 release we will party!

But when AS 7 finally comes out why should you be bothered, especially if you're a user of one of our current releases? There are many reasons and some of them you would probably expect us, or any vendor, about to release the next shiny new product to talk about. So I could talk about the significant performance improvements we've made across a range of projects, such as HornetQ, JBossTS and Hibernate. I could also mention the reduction in runtime footprint that enables us to run the application server on a plug computer. Of course you probably already know about the rapid boot time and additions to our eclipse tool suite that make it one of the top Eclipse downloads. And you may have experienced first hand the developer-oriented features such as the new logging or streamlined build processes. So I'm not going to mention any of those things!

Instead I'm going to concentrate on a few of my personal favourites as well as those of the rest of the JBoss team. So in no specific order, here are 10 things you really need to know about JBossAS 7:

1: full and web profile implementation of Java Enterprise Edition 6. Of course we've been heading in this direction for a while, with previews in JBossAS 5 and a web profile in JBossAS 6, but this time we'll complete that journey and in style! Implementing EE6 is a challenge that only some vendors can match, but implementing it in a way that really shows the full power and flexibility of the standard is a far better and meaningful approach! Several of the key components within our application server have either been rewritten or updated to become best of breed, such as HornetQ, RESTEasy or Infinispan, or were best of breed to start with, such as JBossTS, Web Services or JGroups. If the whole is greater than or equal to the sum of its individual parts then this isn't just any old EE6 implementation, it's the best EE6 implementation!

2: a core part of EE6 and one of it's greatest improvements over previous versions, is CDI. This is a specification which we lead and drove through the JCP and where we also provide the reference implementation, Weld. The specification grew out of our experiences with Seam as well as various legacy frameworks that rely on XML for bindings. CDI makes the development of complex applications straightforward because it hides all of the complexities of non-functional capabilities such as transactions and security. And importantly it does this through annotations, so that you develop using POJOs, which is far more natural for Java developers than early 21st century techniques!

3: a new modular services container replacing the JBossAS 5 & 6 microcontainer. Now before you roll your eyes and move on to the next item, I know better than most that modular containers have been around for a long time. The original Bluestone Total-E-Server/HP-AS was one of the first implementations and we even created a JSR called the Core Services Framework to try to standardise on this type of architectural approach. And JBoss has had a couple of goes at this too. But this time we went back to basics, with the Modular Services Container, and wrote new algorithms for checking module/service dependencies at deploy and run time, such that only the services you need are loaded and started, and when no longer needed unloaded. You no longer have to trim your deployment to get a minimal memory footprint and start time: we give you them out of the box, and if you add your own services later to create new "profiles" then you can be sure that you'll always get the optimal configurations because the container will enforce that for you. As I've said several times over the past year or so, this is very much a SOA-like approach.

4: no more classloader hell! In conjunction with the MSC redesign, the team also took the opportunity to rewrite the book on how classloaders work. Now how you use classloaders, what type of classloader to use and when, are much more carefully defined. The modular approach makes the scoping of classloaders easier and more narrowly defined, helping to avoid conflicts: now you have to explicitly allow deployed services and applications to see and use dependencies. Oh and as a result of this rearchitecture you'll also see a performance improvement as the application server no longer scans the classpath every time a new class is loaded.

5: testable by design. Over the years we've learnt a lot about testing our projects in isolation and in conjunction with each other. Of course we use Hudson/Jenkins and we also inherited the Distributed Test Framework. But it wasn't really until the development of JBossAS 6 & 7 that we had the need to make sure that they were implemented from the ground up with testability in mind. As a result we have Arquillian which is being implemented in lock step with JBossAS 7 and also proving to be a hit in it's own right. In fact the combination of JBossAS and Arquillian offer an environment that is unparalleled in it's simplicity and power. Factor in other projects, such as Byteman, that allow for far more complex testing, such as fault injection, and you have an environment that will produce the most tested implementation yet.

6: configuration and domain management has been improved significantly. Gone are the days when you had N different configuration files for the M different components within the application server. Now there will be only one! And changing configurations is much better supported through a shell (command line) and programmatically, with both fully synchronised. If you've used Un*x shells then you'll find familiarity with the shell, with tab completion and an extensive suite of commands, which allow you to monitor and manage the application server as well as applications that are deployed on it. But beyond making managing a single instance easier, we've gone much further with clusters. Manual updates of changes to individual cluster members, often an error prone approach, have been replaced with the new domain manager, which automates the process using a master/slave or coordinator/cohort mechanism.

7: RESTful services and interfaces throughout. Of course with the adoption of JAX-RS as part of EE6, REST is now as well supported and available to application developers as Web Services. But that's not the same thing as having core parts of the application server available through REST. In JBossAS 7 we've now got projects such as HornetQ, Infinispan and management, exposing REST interfaces. And other work is being driven through the RESTEasy project as REST-Star, including transactions and security.

8: OSGi has evolved a lot over the years from its narrowly scope beginning. These days it seems that everyone is either looking to implement OSGi containers or develop their applications using OSGi bundles. A while back we announced that we would support OSGi on EAP 5 and beyond, but that we would not be using one of the containers already available. The reasons for supporting OSGi are fairly obvious, including the fact that it has become a de facto standard and we want to continue our mission of portability of applications (plus for some scenarios, such as SOA, a bundle is a very natural choice for deploying a service). But why not use one of the existing implementations? Well for a start we've got our own excellent container implementation. But realistically we want to support a range of models, so we don't want to rule anything out at this point. And our OSGi implementation on the new JBossAS7 container has improved the support significantly.

9: multiple language support. Ok, not quite a feature of JBossAS 7, but because we've made sure that we focus on developers and what they want, we've opened up a whole range of use cases that were rather difficult to do previously. So projects like TorqueBox, which offer Ruby users access to enterprise capabilities, have been able to spring up on JBossAS 6 & 7. And other projects, such as Infinispan, are finding calls for integration with, say, Scala on the rise. This is great, because it helps to showcase JBossEverywhere once again!

10: We've been sure to make the architecture of the application server cloud-friendly, so that it's the perfect substrate for an enterprise Java PaaS (and through indirection and abstraction techniques such as those employed by TorqueBox, for a much wider choice of languages as well.) Of course it's going to be a number of years before the Java Enterprise Edition standard really tackles the cloud through EE7 and EE8, but we are already pushing that envelope. If you're interested in seeing how the future of PaaS will unfold then you can get a glimpse of that by watching JBossAS 7 and beyond.

So there you go. 10 things that make JBossAS 7 stand out from its predecessors and the competition. Hopefully when you try it out you'll agree with some of these and maybe come up with some of your own (some things we may take for granted might be significant improvements in your life.) Over the next few weeks you'll see and hear more from us on AS7. And maybe, just maybe, I'll get a chance to make some more bigger announcements soon!