XForms Everywhere

We just committed changes in Orbeon Forms to handle a disk-based XForms state store. What is this about? As discussed recently, Orbeon Forms needs to store XForms state information somewhere. One option is client-side, but this has drawbacks in terms of bandwidth and latency. So the default since the beginning has been to store state on the server.

Until this last round of changes, Orbeon Forms would keep some of the information in a session LRU memory store and then discard it completely. You had to discard information, because RAM is limited. But this approach has proven problematic because, however good the expiration heuristic, in many cases the state information would be lost completely, leaving the user in the dust. This includes cases such as the user going back to a page after a while, or accessing multiple large pages within a same session.

So we had to fix this, and you can say welcome to the brand new disk-based session store! With this mechanism, state information which expires from RAM is instead migrated to a disk store. Since disk is cheaper than RAM (you can maybe allocate a few tens of MB of state store in RAM, but probably several GB of state store on disk, at least two orders of magnitude more), memory is no longer really a constraint. (This can work because disk access is expected to be rare, following the principle of locality of access, which is central to all caching systems.)

Besides making the user experience better, we hope that this larger and more reliable store will enable Orbeon forms to support very cool features soon, including built-in undo and history checkpoints in forms. This should take Orbeon Forms where few Ajax platforms have already been.

Even with a disk-based store, you need to expire data otherwise it will grow indefinitely. In this first iteration of the store, we implement a fairly simple expiration stategy: stuff expires from the database when the user session is terminated. This is not perfect, but will cover a large number of uses cases.

Technically, we have made a fairly bold step and decided to use the embedded eXist database, included with Orbeon Forms, as the persistence layer for XForms state information. One reason is that we did not want to deal with writing our own persistence layer. Another reason is that this will give us the power of XQuery to implement expiration strategies in the future.

This feature is now available in the Orbeon 3.6 beta nightly builds, and will be enabled by default in Orbeon Forms 3.6.

We will speak at the XForms evening that will take place alongside the XML 2007 Conference in Boston (3-5 December 2007). The XForms evening will take place on Monday, December 3 and will feature seven 15-minute talks (we want to make sure you won’t get bored). Our talk will be about XForms and eXist:

XForms and the eXist XML database: a perfect couple

XForms speaks XML natively, and so does the open source eXist XML database. In this talk, we show how they form a particularly attractive combination.

In particular, we show how the XForms 1.1 submission module, which supports REST, can be used to perform CRUD operations in eXist. We also look at how XForms can directly submit XML database queries using the powerful XQuery 1.0 language implemented by eXist. The result is an architecture where complete applications can be developed with an XForms processor and the eXist database, while relying entirely on open standards.

We conclude the presentation with live demonstrations of the XForms + eXist combination built using open source software.

We hope to see your there!

I had a chance to work with a great team on the new edition of Professional XML. This is a completely new edition; everything is fresh and has be written from the ground up. The book covers almost all the XML-related technologies you can think of: XSLT, XPath, XML Schema, Relax NG, XSL-FO, XHTML, XQuery, Ajax, RSS, Atom, Web services, and many others.

For the XPath chapter, instead of writing a 30-page overview of XPath, I decided to take a different approach: looking at a number of specific features of XPath that are both very useful in practice, and somehow lesser known or not obvious. I have already alluded to some in this very blog:

• Debugging with trace()
• Reverse Axis, Evil at Times
• The unordered() function, Quite an Oddity
• Tuning your XPath Expressions
• The Many Faces of a Document
• When A != B is different from not(A = B)

Grab a copy of the book, and you will find more about XPath, and of course much more about a number of other XML technologies. I hope you will enjoy it!