As usual Peter does a wonderful job explaining programming concepts. He has a calm and natural way of explaining things verbally that works very well for this kind of presentation (or for a podcast interview).
I was certainly impressed with LightWire. As he says, it may not be for everyone, but it makes a good alternative to ColdSpring. LightWire seems to have most of the features of ColdSpring, but not all. For example, ColdSpring supports AOP which LightWire doesn't (yet, Peter indicated plans to add that later).
LightWire, however, presents some advantages of its own. It has fewer files and doesn't need to be in a specific location or use mappings as ColdSpring does. Additionally, you can configure LightWire via the same XML syntax as ColdSpring or via a programmatic API.
Personally, I find the choice very nice. I can certainly see scenarios where each are advantageous.
Right now I have my own DI/IOC Engine. Mine was created before ColdSpring or LightWire and isn't as robust as either of them. I am likely to migrate to one of them (especially if a needed feature or two is added) in the near future.
When that time comes, LightWire certainly seems like an attractive choice.
As an aside, I have noticed an attitude by some developers to disparage LightWire for daring to challenge the "One True IOC Framework". This attitude is disappointing as more options seems only to the benefit of all of us.
Certainly, I think we are better off now that we have more than one MVC framework. It would have been a pity, for example, if Joe Rinehart hadn't created Model Glue merely because of the existence of Mach-II.
Anyway, it was a really good presentation and well worth the listen even if you have no plans of trying out LightWire. Peter always has good explanations of his thinking. The nature of the problems that he is trying to solve (and the degree to which he researches them) make his a valuable viewpoint to hear.