brentashley

I’ve been following advances in the Ajax world so I can keep my Ajax Transport Layer Alternatives presentation up to date for The Ajax Experience in Boston next month.

Harry Fuecks wrote recently about new approaches to Javascript asynchronous calls. It’s really neat stuff and I’m stretching my brain to try to understand it fully. I wonder though to what extent it will actually solve problems that really exist for most people any better than what already exists.

I’ve been making Ajax apps that work well enough for years while many people have been waiting for all the stars to align before they even try it. I’ve had all sorts of flack from pedants for using iframes and img/cookie because they’re hacks, but JSRS and RSLite have worked consistently and predictably across a large number of browsers for 5 years and more without modification. I only changed my Blogchat app to use XMLHttpRequest recently (for no really good reason – it’s been unchanged since 2002) and the first thing that happened was a huge debugging session to figure out a really wonky deep IE7 issue.

The thing about simplicity in the programming world is that it begets stability. The unknowns and dependencies introduced by layers of abstraction and frameworks and preprocessing can in some cases introduce far more potential complexity and maintenance issues than the problem at hand is worth. While there is definitely a class of complex UI problems that are now trivially resolved using the latest and greatest of libraries and frameworks, don’t forget that many simple problems deserve simple solutions.

Those of you who visit my blog with a browser rather than a newsreader may have noticed that for the past couple of weeks I’ve had a new banner up for the new Ajax Experience show in Boston on October 23-25 2006.

I’ll be there giving my Ajax Transport Layer Alternatives presentation, participating on panels and mixing with the attendees and presenters.

The first AE show was in May this year and was a resounding success for both the experts and the audience.

As I said to Ajaxian founder Dion Almaer when accepting the invitation for Boston, I’m really interested to follow up on the potential advances sown at the last AE and see whether I (have managed to || continue to) exert any influence whatsoever as regards connecting people who need to work together to make this technology realize its potential.

Again the sessions are 90 minutes each, so lots of time for Q&A. Although many of the sessions from the first show are slated to be repeated (many like mine updated with current information), with 5 concurrent tracks, if you attended before this is an opportunity to get to see the ones you missed last time.

The swag was great at the last show – branded AE iPod Shuffles for all. I understand alumni will qualify for a Nano this time!

If you come to the show, be sure to find me. I’ll be fairly easy to find wearing my Ajax duds.

Pete Forde reinvents search with his Live Filter demo.

Take particular note of Pete’s never-ending auto-loading page that gets more elements from the result pool as you reach the bottom of the page. Try it yourself – click the View Results button right at the start to see all 64 items, then scroll to the bottom and watch it get the next set of items on demand before you even knew you were demanding them.

Pete and the Unspace folks have been taking serious aim at improving the state of UI interaction for the good of us all. Not content with simply implementing something like OpenRico’s LiveGrid, they apply even more anticipatory design to fixing the scroll-within-a-scroll UI anomaly that LiveGrid introduces to a page that uses it.

The result is that by anticipating a need I didn’t even know I had, they have provided an interface that is natural to use yet very powerful.

I’ve been reading the IE Blog which the IE development team hosts. I’ve started to notice that every time I see an entry it’s someone else and they’re all “Program Managers”. Is EVERYBODY a PM at MS? Is there nobody that actually DOES anything? Sounds like layered middle management run amok. From the names and titles I can see, IE seems to be divided into so many separately managed projects that they must have program-manager-managers that just coordinate information flow amongst the program managers. Yeesh.

I’ll be at the Bodega on Westgate Road near the Central Station in Newcastle at 5:30 or so on Thursday evening Aug 10th, having a yap with some of the local Newcastle web development community. If you happen to be in the area, be sure to drop by.

I was at DemoCamp 8 last night and as usual stayed later for the social networking. I got into a discussion with Pete Forde and others about who does what.

I have real trouble explaining to people what it is that I do. I often try to explain the nuance between being a scripter (which I am in spades) and a programmer (which I am not exactly) and the distinction falls short of conveying the gestalt that is the technical side of me.

I spend a lot of my time making a whole mess of things work together that were never meant to be integrated. It often takes a wide and deep range of knowledge to figure out the bits. I have variously called myself a technical spot-welder, a duct-taper, a spelunker, and a special-ops data diver, trained to get in under the wire and get out with the data.

Here’s an example – at one of my clients, something we do is give realtime tests against devices in the field. A diagnostic page could tell you, for instance, the contents of the ARP resolution table on a remote router. There is a bunch of magic that goes on behind the scenes to do this (simplified here):

• the user presses the “get ARP table” button
  • the external portal web server (ASP.NET/Win2k3/MSSQL) makes a SOAP call to get the information
    • the soap call is received at an internal webservices server (Apache/PHP/BSD)
    • connection information is resolved against a PostgreSQL database
    • a shell script is invoked on the webservices server
      • the shell script uses ssh to connect over a secure tunnel to an API server at the customer premises that can route to the destination
      • a script is run on the API server (bourne shell, perl, telnet, curl, grep, sed, awk, lynx etc)
        • the script connects to the router, logs in, displays and parses information, logs out
    • info gets routed back to webservices server via the ssh connection
    • php assembles SOAP response
  • ASP.NET receves SOAP, builds result page
• Result received by user

I make all that stuff happen.

Now how do I distill that into an elevator pitch?

Last week I received a complimentary copy of fellow Torontonian Peter Lavin‘s new book, Object Oriented PHP, published by NoStarch Press. I’ve had a couple of really good books from NoStarch (Absolute OpenBSD is indispensable if you use that OS) and this is no exception to that trend.

The book introduces Object Oriented programming to PHP programmers, starting with PHP4’s OO implementation and demonstrating the advances made with PHP5. A simple web application is built by starting small and making incremental changes, sequenced in such a way as to introduce OO concepts and solutions in a natural way, increasing in utility and complexity as the lessons progress.

I’m only a few chapters in right now – there is certainly more to the book and I’m looking forward to the rest.

The companion website has a lot of useful reference material and links to resources, as well as the book’s code and errata.

(via Ajaxian) IBM Developerworks has just published an Ajax Transport Method tutorial that describes a few of the alternatives to XMLHttp and provides example code to make them work.

For a more complete reference to all the available Ajax Transport layers, see my constantly evolving Ajax Transport Layer Alternatives presentation documentation. As someone who has actively used and promoted these techniques since 1998, I can say that I’m pretty familiar with the entire gamut of methods and their pros and cons.

The universal acceptance of the Ajax meme only became possible after more than 7 years of evolution and convergence of tools and techniques – CSS and DOM for presentation, XML and JSON for encoding, and various transport layers.

Much of the experimentation that took place during this time was centred around finding a way to transport data to and from the server in the background. Java applets, frames, layers, hidden iframes, image tags, script tags, XMLHttpRequest and Flash were all put to the test, each solving parts of the problem but mostly failing on compatibility, reliability or unwanted sideffects. XHR’s failure to take hold for the first 5 years of its existence was due to it being implemented as an ActiveX control only available in IE.

When Mozilla, Opera and Safari included the XMLHttpRequest object, it finally became the “good enough” solution upon which the Ajax juggernaut was launched.

While XMLHttpRequest has served to bring us to this stage in the game, it has some serious shortcomings that limit its long-term utility. As far back as 2002, Scott Andrew LePera pointed out an as-yet unfixed implementation problem, and more recently, Alex Bosworth enumerated a whole raft of issues. Harry Fuecks has spoken eloquently about network reliability and latency, the solutions to which would be best addressed with better control over timeouts, retries and error handling than XHR can provide.

As I’ve mentioned many times, Douglas Crockford’s JSONRequest proposal is one of the first real attempts to begin the discussions necessary to bring the changes about that will solve these issues. There was futher discussion at The Ajax Experience that included the major browser vendors, so I’m hoping that some from these beginnings, momentum will build.

Until there is a purpose-built Ajax transport built into browsers, advanced developers may want to learn all of the ins and outs of the different alternatives in order to choose the one that best fits the problem they are solving.