The much simpler idea is to stop using bloated, general purpose frameworks, and just use tiny, functional-style hooks when needed, the way people used JS in days before the framework madness.)

After using several frameworks in various projects, I must agree with that. The main problem with frameworks is that they work well until you try to do something they weren't designed for - and then you end up with a pile of hacky and messy code. With smaller libraries it's much easier to replace one for another or just write your own small piece of code that will do exactly what you want.

Could you please go more in depth about "functional-style hooks ... before the framework madness"? I'd be interested in reading your thoughts as a newbie JS dev currently studying backbone/angular/jquery and other libraries.

I believe that what he meant (Correct me if I'm wrong), is that instead of creating a system of classes that tries to abstract everything out of the DOM, it's better act on the DOM itself, via jQuery, Zepto.js, etc.

In some ways, what I showed in the post is functional.

- Select a DOM i.e var $elem = $("#anElementId")

- Act on it i.e $elem.fadeOut()

And yet what about the "JQuery soup" problem that everyone complains about? Isn't that due to having DOM manipulation code strewn about willy-nilly?

Sometimes, it's good to encapsulate stuff into a class when you have a complex logic around a specific theme.

It's not one or the other. It's one AND the other.

There was a nice set of slides few days ago - http://www.slideshare.net/nzakas/enough-withthejavascriptalr... - exactly the same idea.

Of course, the drawback of this approach is that you're now sending javascript with every request. If you had done the traditional approach of sending your javascript in one big minified chunk it only needs to be sent once and your browser will load it from cache on subsequent requests.

You are right, and I quickly went over it because I was scared of loosing the focus of the post by explaining into details many small parts.

This is a balance you have to consider. If a user never clicks on the link, this method has an advantage over minifying everything since you would load the JS without using it.

On the other hand, if a user keeps clicking the link while being on the same page (Important to stay on the same page), it would be better to have it stored in your minified chunk.

It's a low hanging fruit optimization that can be very useful IMO.

GWT actually does this automatically, works very nicely: http://www.gwtproject.org/doc/latest/DevGuideCodeSplitting.h...

That is very cool, I'm reading it right now, thanks for the link

Here's pothibo's talk code samples that he mentions in the beginning of the post: https://github.com/pothibo/open-code-ajax

I liked the idea and added a demo using Flask/Python as well :)

Thanks for sharing!

The talk is in french so you guys might want to skip the talk and jump to the code ;)

I've always wondered if it wouldn't be interesting to treat certain parts of a web application kind of like Objective-C NIB files. NIBs are sort of used to load new code into the application at run-time. We could do something similar with javascript maybe? Load parts of the UI in chunks based on what the user is currently doing with the application so that initial load times are lower.

Yes I believe this could very well exist and I'd be happy using something like this. I thought about using WebSockets to send HTML/JS stuff down the pipe as the user navigates. Would cut some lagging due to how HTTP request are made and would decrease the load time.

NOLOH (http://www.noloh.com) has been doing this automatically since 2006.

Disclaimer: I was one of the co-founders.

This is actually a step up from what I've been doing which is using views to selectively load what JS a certain page actually needs - instead of bloating it up with several different libraries or frameworks that are all gang banging the page when it loads.

Performance wise, I've seen a significant increase in page load times.