IFCX : Internet Foundation Classes eXtreme!

IFCX Wings v0.7 is now available from the usual place (wow, spiffy new site design for SF.net!). It features an in-progress (during script execution) console that pops up during long running scripts. Very handy to see that something is happening while Ivy downloads your JAR dependencies. Default support for OCaml, Scheme, and Adenine has also been added.

I've tested the Wings.odt with OOo 3.0beta2 on Windows XP and Mac OS X Leopard and it seems to work fine except that on Mac there is no support for AWT (and consequently no Swing). It also works with OOo 2.4 on Windows and NeoOffice 2.2 on Mac. Earlier versions are mostly OK too, but don't go toooo far (like 2.0) or GroovyForOpenOffice won't work (and you need it for Wings).

Note that you must install Groovy For OpenOffice for the WingsEval macro to work. Otherwise you can just open the Wings document and read it, but the code won't be executable.

The keyboard shortcut for the Wings macro seems to work with all those versions and platforms too. The Wings.odt file is signed so that you can have some reasonable expectation that we're using the same bits. Of course if yours is not signed by my CA Cert showing JAMES PAUL WHITE and my email address then there may be a problem.

OCaml (Objective Caml) is a popular version of ML sporting OO features. OCaml-Java is the spiffy implementation for the JVM by Xavier Clerc used in Wings.

Adenine is the language of the MIT Haystack Semantic Desktop and is essentially a LISP for graphs. It features a RDF data model, Pythonic syntax, and is implemented for the JVM. I have extricated it's implementation from Eclipse and packaged it with a JSR-223 engine adapter so that Wings can support it by default. The Adenine Tutorial converted to IFCX Wings is available from SVN.

The Scheme support is using the SISC implementation (even though I'm long time Kawahead) because it has a JSR-223 engine already and it is also better for pedagogic purposes since it has full continuations.

There is also exciting news regarding IFCX.org and the OpenOffice.org Community Innovation Program, but I'll hold off on the details of that until November.

The v0.6 release of IFCX Groovy For OpenOffice and Wings (version numbers bumped to align for the moment), demonstrated at CommunityOne and JavaOne last week, are available for download in the usual place.

A big THANK YOU to Brendan Humphreys and Peter Moore of Atlassian for the invitation and support to present Wings in a lightning talk at CommunityOne. I did have a nasty demo devil bite that prevented a live demonstration, so next time I'll have a Time Machine backup with me...

This release of Wings should be useful on at least an experimental level for the adventurous. The examples include plenty of Groovy of course, but also Ruby/JRuby, Python/Jython, and Haskell/Jaskell.

Apache Ivy integration enables JAR dependencies to be loaded dynamically and languages are pluggable via the JSR 223: Scripting for the Java Platform. JDK 1.5 is all that is required, although JDK 1.6 or 1.7 are fine of course.

G4OO v0.6 features Groovy 1.5.6 and Apache Ivy 2.0-SNAPSHOT bundled with the Ivy RoundUp Builder Resolver.

Using that pluggable language scheme, and thanks to a sprint at the Scala lift off last Saturday with the invaluable assistance of Lex Spoon and Toomas Romer, WE HAVE SCALA! You can download it from the usual place.

A recent post on the OCJUG list asked to what degree is Java "open" and when will everyone agree that it is "open" to their satisfaction.

The fact that "Java" is a trademark (the significance of which was made obvious when Sun changed their stock symbol from SUNW to JAVA) means that what most folks think Java is will never be as open or free as some folks want.

Furthermore, Sun has a dual license for their Java distribution which includes OpenJDK. So if you want to contribute to the most popular JDK distribution, you have to agree that Sun can also use your efforts under terms that are not necessarily free or open (although they do promise to always make your contribution available under a FSF and/or OSI approved license).

http://openjdk.java.net/contribute/

So OpenJDK has two strikes against it.

Here comes the next pitch, aaannd OpenJDK is GPL licensed, a home run!

I was amazed and pleased that Sun went all the way and used GPL for Java. After all, I'd been saying that Java would eventually be OSS, even after the ISO talk was replaced by the JCP.

That means this pillar of the Great Java Renaissance of 2006 is strong as it possibly can be. One of those strengths is that anyone who isn't satisfied with the degree of openness can simply fork with total abandon and with the best license for software freedom. That is a reality and one of the first projects spawned by OpenJDK is IcedTea, which is a free and open implementation of Java.

As for folks saying Java being OSS doesn't matter. They are quite mistaken. Already we've seen significant developments such as SoyLatte which counters Apple's weak support of Java, and research work on Java being truly free and open in the Da Vinci Machine Project. This change in research work is important because in the past such work, if done with Sun's JDK, used a license that meant the resulting code rarely ever left the university.

That the OpenJDK is an effective OSS project is clear because all of these efforts are now part of the OpenJDK project itself, rather than forking or otherwise choosing to stand alone or with another group.

That is only the beginning and, by being real Open Source Software, we can rest assured that Java will grow in strength over the next decade just as it did the first. Of course we can also be assured that there are plenty of folks that will disagree with just about every aspect of all this jazz.

Alex Tkachman posted a question about Groovy style for checkpointed calculations.

Here's my first try:

// Proof of concept for Groovy checkpointed calculation with closures.
// @author Jim White <jim@pagemsiths.com>
// http://www.ifcx.org/

Session session = new Session()

// def initState = new Expando(data:'datalocation')
def initState = [data:'datalocation']

def endState = eachWithCheckpoints(session, 'state1', initState, [
      { stepCount = longInitialConditionsCalculation(data) }
    , { accumulator = 0
        iterateWithCheckpoints(session, 'state2', it.state
           , 0..stepCount
           , { accumulator += 2 * it.step })
      }
    , { result = "We did $stepCount iterations and the answer is $accumulator!" }
])

def longInitialConditionsCalculation(d) { 10 }

println endState.result

class Session {
    Object loadCheckpoint(String k) { null }
    void saveCheckpoint(String k, Serializable s) { }
}

def eachWithCheckpoints(Session session, String key, def initialState, List<Closure> closures) {
    def control = session.loadCheckpoint(key)

    if (control.is(null)) {
        control = [step:0, state:initialState]
    }

    while (control.step < closures.size()) {
        Closure clos = closures[control.step]

        clos.delegate = control.state
        clos.resolveStrategy = Closure.DELEGATE_FIRST

        clos.call(control)

        control.step += 1

        session.saveCheckpoint(key, control)
    }

    control.state
}

def iterateWithCheckpoints(Session session, String key, def initialState, Range range, Closure clos) {
    def control = session.loadCheckpoint(key)

    if (control.is(null)) {
        // Leave reversed range case as an exercise for the reader...
        assert !range.isReverse()

        control = [step:range.from, state:initialState]
    }

    clos.delegate = control.state
    clos.resolveStrategy = Closure.DELEGATE_FIRST

    while (control.step <= range.to) {
        clos.call(control)

        control.step += 1

        session.saveCheckpoint(key, control)
    }

    control.state
}

==>

We did 10 iterations and the answer is 110!

The idea is either you're doing a sequence of different steps or iterating the same step some number of times. Notice that the state can be any serializable thing. Map or Expando is handy, but some bean or other class would be fine.

Obviously there are further refinements possible such as reversed and stepped ranges, and also a more functional style is possible. Either a builder or controller class would streamline things a bit by hiding some of the details of the eachWithCheckpoints/iterateWithCheckpoints calls.

If Groovy had serializable iterators for the control step, that would make iterateWithCheckpoints nicer and let it change from taking just a simple integer range to a list.

Naturally where there this is headed is cloud-powered click-and-it-goes Wings via your web browser using OOHTML.

I've been interested in TREE-META since I discovered it while at UC Irvine. We used a version for the UCSD p-System on the Teraks (desktop LSI-11/23). It had a pleasant and compact syntax that was more convenient than the LISP that took it's place for me.

For several years I've been on the lookout for information on TREE-META but haven't been able to find out a great deal. After about the third time listening to The Mother of All Demos I heard them mention that TREE-META was the language used to implement the "special purpose languages", which we call "domain-specific" today.

I've been unable to locate a copy of any version the program or the tech report describing the language, although I've seen hints than some folks have used it fairly recently.

In commemoration of the 40th anniversary of the MoAD (also at Wired), I've created a Wikipedia stub about TREE-META as well as one here for non-WP suitable material. The WP article is very rough, but I'm hoping that others with information will bring it forward and collect it there.

Got my G1G1 XO Laptop yesterday evening and have been fiddling with it for it hours. I'm very pleased and impressed with the design and believe that the XO will deliver on OLPC's vision of bringing computer literacy to, and unleashing creativity in, the children of the developing world. Actually I'm convinced it will do that for more folks than just them!

I did take some pictures of course, but I'll point you to some videos that do a better job of introducing the XO:

• Nice review by David Pogue
• Sound bites from OLPC team
• Brief tour of Sugar

Sugar is the XO's desktop GUI. It is very good and has some excellent innovations. Three of the elements that stand out are visual map of the mesh network neighborhood, visual map of running applications (the "activities" in the circle are running, as opposed to the menu bar on the bottom which are those that can be launched), and, most importantly IMO, the activity-centric Journal (rather than the expired files-and-folders desktop metaphor). Having a task-oriented organization scheme is something I've been wanting/expecting in Mac OS for twenty years, perhaps now it'll happen when the Apple folks see they've been scooped in UI design by a Linux machine with 128MB of RAM (it's 1984 all over again!). Of course Genius Folders will be a dandy enhancement to the Journal's tagging scheme.

For young children the XO is probably darn near perfect. The built-in tools (writing, music, chat, web browser, video, graphics, data recorder, and calculator) and games are the obvious starting point. For real computer powered creativity it has an impressive set of easy-to-use programming tools. Python is the XO's primary scripting language and the Pippy and Develop (Activity-building Activity) IDEs are a natural progression from EToys (a LOGO-like environment in Squeak/Smalltalk) and Turtle Art visual programming activities.

Some obvious low-hanging fruit for making the XO a vehicle for teaching older children and young adults is packaging eBooks and on-line course material from sources like Project Guttenberg, MIT OpenCourseWare, and Stanford on iTunes U. I see a flourishing library system of SD cards and USB memories. Naturally the same materials would work dandy on the Sony and Amazon eBook readers. Somewhat more challenging (but of personal interest to me) is a web newsfeed system to replace my newspaper subscription.

Java-oriented folks will of course be disappointed that XO doesn't normally support Java, but being a resource-constrained platform that is pretty much unavoidable (Microsoft has been making the same whine wrt Windows on XO). While there has been some work done to make Java available on XO including a JNLP handler, I think a more useful approach would be Google Android for XO. And I'm sure Gosling would agree since it is his opinion that the PC for the developing world is the cell phone and that the OLPC Project is bad idea.

Another idea for developers I have is porting Sugar to the Nokia N800/N810 which have specifications that are similar to the XO (actually they're a bit slower and ARM-powered but the memory and display sizes are quite close). There is some discussion about Sugar and Nokia's Maemo but there doesn't seem to be any indication of a port in progress.

There is still time to participate in Give One Get One as the deadline was extended to December 31st, so if you haven't ordered yours yet please do it now! Not only do you get a tax deduction for the donated XO, you also get a year of T-Mobile WiFi which covers a lot of places including a zillion Starbucks.

When you do get your XO Laptop, please let me know as I'm interested in meeting up with other OLPC-minded folks.