Steffen Meschkat, Google (worked on Google Maps)
========================

Reality-Checking the Ajax Web Application Architecture

Ajax: A bad name is better than no name. It's client-side scripting done the
right way - closely connected to the user experience.

Design patterns showed that names are useful - so we'll stick with Ajax.

In classic web applications you have forms and pages. Everything happens on
the server.

Consequences of the Ajax model:
* Sophisticated user interaction (like 1990!)
* Client side session state - transient session state is managed on the
  client, while persistent user state is maintained on the server. This
  corrects a long standing architecturlal aberration. You don't need
  server-side session state any more!

Sessions also cause problems with scaling - you need your whole cluster to
know about the session data.

Intermission:

The bad thing about doing something right the first time is that nobody
appreciates how difficult it was. Web technologies give us plenty of
opportunity to appreciate how difficult it was.

I didn't quite follow here, but I think he was criticizing web development as
a flawed environment.

He slates CSS, DOM and JavaScript as all flawed in various ways. DOM nodes
have "unspecified partial deserialization semantics" - it is unspecified what
the DOM looks like if a page is only half loaded. Also has inconsistent naming
schemes. These dumb errors are representative of the rushed nature of the
specification.

JavaScript - semicolon insertion is bad. Single threaded, but no one tells you
so. So, if your JavaScript takes a long time to execute your browser freezes
up.

Prefers JSON to XML.

Practical consequences:

Cross browser compatibility problems enforce good libraries. Separation of
interaction logic and application logic (they're even written in different
languages) is a good thing.

Event registration:

function ev(node, name, handler) {
  if (/safari/i.test(navigator.userAgent) && event == 'dblclick') {
    node['ondblclick'] = handler;
  }
  // rest is as normal
}

Despite the suckiness of JavaScript (stupid name, semicolon insertion,
surprising scope rules, poorly specified runtime) it has some cool features:
    Custom objects, delegation, closures
    Rich literals, exceptions, functors
    Has been called "Lisp with C syntax"
    The name made sense at the time (for scripting Java applets from HTML
    pages)

Demonstration involving a closures for Google Maps. Had a memory leak in IE -
it's possible that his next bit (which he had to cut) would show how to get
rid of it using a factory function.

Challenges:
    Deployment (caching, compilation, modularization)
    Bookmarking and History
    Graceful degredation
    Frameworks