Monday, June 13, 2005

I was just doing some research for what seemed like a pretty simple, benign, blog posting, and wound up spending two hours following links on Ward Cunningham's wiki...

I figure it might be helpful to share some of the highlights of my scouring.

Along with a few other XP terms, the term "Unit Test" is in the process of "officially" being changed to "Programmer Test" apparently in an effort to avoid continuing confrontations with the very powerful and demanding Quality Assurance Industrial Complex.

Apparently, shockingly, I'm not the first person to hope for, look for and dream about the most amazing programming language "ever"...

(That's three different links you should click on)

I found out that the thing I tend to do when nothing logical seems to be working in Flash, actually has a name.

I also learned that if you're using blogger.com, and you're editing a blog, CTRL+S actually saves AND publishes the page you're editing - which subsequently sends an email to anyone that is subscribed to your blog.

Sweet.

Tuesday, June 07, 2005

Just wanted to give a heads up...

Ali and I had a talk with Robert Penner last Friday to discuss his involvement in ongoing AsUnit development.

As it turned out, he had quite a few really GREAT ideas about how AsUnit could be improved.

One of them was that we could batch our Local Connection transmissions so that the UI would be more responsive.

I was promptly up until the wee hours last night trying to get it to work. Knowing what I know about the implementation, I figured that it would be a pretty easy addition...

As it turned out, I was able to get some small tests working, but as soon as I ran some of our bigger applications, none of the Tests were being received by the server. There was no call to onStatus, no obvious reason for the failure, no notification whatsoever. Eventually, (around 1:30am or so) I figured that I probably shouldn't be coding and drinking and figured I must be doing something stupid!

This morning, I caught a surprising conversation on the osflash list - they were talking about how the Local Connection object has a 40k data limit! What?!

So I checked the implementation and if I put a 100 assertion limit on the batches, everything worked fine - even on the big builds.

So -

I dropped a build on AsUnit that has a 100 assertion batch size, but this is kind of arbitrary - especially considering that folks could put just about any size of string message into each assertion and even more so because I usually don't put anything in the message field.

I now want to put together a build of LocalConnClient and LocalConnServer that has some kind of handshake mechanism, so that any tranmission also includes confirmation of receipt. If the server doesn't confirm, the client will throw some kind of error. We can then handle this error in AsUnit by cutting the batch size down a bit, and retrying the request.

I figured folks ought to know about this - since it seems to be pretty important if you're building an app that relies on Local Connection...