It is not possible to inject spring managed beans into a JPA EntityListener class. This is because the JPA listener mechanism should be based on a stateless class, so the methods are effectively static, and non-context aware.

In Hibernate at least this is not strictly true, since if you were (and you shouldn’t) to record an entity in a field of this in a method, it’s sat there next time the method comes along.

public class EntityListener {
   private Entity entity;

   @PostUpdate
   public void doStuff(Entity entity) {
       // don't try this at home, it is a BAD THING
       this.entity = entity;
       ...
   }
}

This is more than a little daft, what with the lingering reference, and the concurrency nightmare that would no doubt ensue.

No amount of AOP will save you, nothing gets injected to the ‘object’ representing the listener, because the implementations don’t actually create instances, but uses the class method. This is a perfectly reasonable approach based on the JPA spec, so is not really a bug, but does mean that there is no bean to inject dependencies to. If you were crazy, you could pointcut each method of your listener, but by that point, you’ve probably lost the plot you were pursuing in the first place, and you might as well just do the whole thing with a pointcut on your target entity (this is debatable, but I will not debate it just yet).

This is certainly the case for Hibernate, and appears to be for at least two of the other major providers (Open JPA, eclipse – ok, so I tested the effect, but didn’t read all the source code for those, because life is too short, and I’d already spent too long reading most of Hibernate).

Of course you could provide access to services through static methods, or a singleton of some sort, but we’re all convinced that’s an anti-pattern, right?

Anyway, this post is running out of time, and will be updated when I get round to cleaning up the rest of the test code.

Here is an account of some interesting experiments in using very sensitive microphones, or voltage meters to crack RSA.

So from this, it would be perfectly possible to build a plug in device which stole anything, even anything encrypted with a key held in RAM, given long enough to listen.

Scary stuff.

I just came across this fantastic article, which has a number of things to contribute. Firstly, how to get that crock-of-browser IE to work.

Secondly, it has some lovely advice for partners on how to deal with us web developers when we’re not going to bed until this works.

Here it is: A Fun Little JSON Murder Mystery With IE9

I thought I’d also point out, partly for reminding me next time I am ripping hair out and depositing it in the gaps between the keys, and partly for the edification of others, that this is far from the only gotcha IE likes to throw up.

Every now and then your find yourself involved with the X-JSON header. This is usually if you’re still using a very old version of Rails, or have for some other reason decided to use a similarly old version of prototype.js. (I assume that if you’re doing this, you’re also battling a scriptaculous addiction – move on, there’s newer shinier stuff). Old version of prototype (1.5 ish) use a custom header to read JSON responses instead of response body, which means you are limited in certain browsers to a 168 characters. This is covered here. One ugly way to deal with this is to have your server pretend it’s sending Content-Type: text/plain, and manually decode the JSON. Or you could upgrade prototype to 1.7, or jQuery.

IE will also usually believe the Content-Length header you give with a JSON response, so make sure you get it right.

Some XMLHttpResquests, and some servers will also get very upset if you send and empty POST request. Always give it some content. Otherwise you probably mean GET, DELETE, HEAD or something like that. If you’re posting something, remember to put it in the envelope, and add the stamp.

Something else to be aware of on the other side of the transaction is dealing with things like Spring’s jackson backed message adapter. Always remember to set the Content-Type of the POST.

Well, I hope some of that will save me and hopefully some others a few late night ‘dur’ moments.

Streaming it live is great. Quite impressed at how well I can watch a conference in Baltimore from the corner of a desk in London and actually feel like its working.

There does seem to be a lot of quotation from ‘popular philosophy’. Do we have to pretend that post modernism has anything to do with web frameworks, just because we read a pithy quote about prepositions, that we thought defined post modernism? Or maybe we could have actually read some post modern philosophy, and wondered what on earth they are talking about. Oh, and inclusive attitude to dependencies and configuration replacing the original essence of convention. I see, well, that’s about as much about postmodernism as Alanis Morissette is about irony.

On the plus-side, I’m rewarming to rails, and while 3 is looking great, I’m interested to see 3.1, and rather hoping that it doesn’t disappear too much up its own behind. After all, contrary to the popular impressions, ruby on rails did not invent everything, and could do with a little bit more of a connection to the rest of the world’s frameworks. Hopefully keeping it anchored in the existing paradigm might just keep it learnable for experienced programmers coming from other worlds.