Reasoning by historical analogy can be dangerous, especially if such reasoning is untempered by recognition that no two historical events are identical and that the future is more than a linear extension of the past. The instructiveness of historical events tends to diminish the greater their distance in time and space from the day and place they occurred.
– PERILS OF REASONING BY HISTORICAL ANALOGY: MUNICH, VIETNAM, AND AMERICAN USE OF FORCE SINCE 1945 by Jeffrey Record
It may seem a little hyperbolic to quote from a paper published by the Air War College in advance of a critique of Joel Spolsky’s latest strategy letter, but lessons about historical reasoning can apply just as well when the question is the future of software as it does when the question is war and peace. So, how exactly did Joel screw up in his analysis of the future of Ajax? Well, first to what he got right:
- He got the bit about the Forms-based interfacre (Green Screen, etc.) right. Web 1.0 was like the mainframe apps, Web 2.0 is like the desktop GUI. Welcome to the club, Joel, over a year and a half later.
- Large JavaScript apps will download faster over bigger pipes (or "tubes" or "trucks") and run faster in newer browsers with better JavaScript runtimes. So nobody cares about the 3k do everything library anymore, just like features are more important that memory footprint and performance in Desktop UI’s. Again, welcome to the club.
- The portable programming language is not likely to be Javascript in the long run. Think VM, like Tamarin or Silverlight. The same developments that make JavaScript run faster in the browser will ultimately make it irrelevant.
So where did Joel go wrong?
- For one, he confuses the categories of Hardware Platforms, Operating Systems, UI Toolkits and Application Software. Anyone remember the RogueWave GUI libraries, circa 1994, or Think’s Lightspeed C for the Mac from way back when? Where are they now? (RogueWave is still around, as is Symantec which acquired Think C, but you wouldn’t exactly say they were the "winners" when it came to GUI SDK’s.)
- What are the applications, the 1-2-3′s of the web age? Well, everyone who publishes a web site is an ISV. Some make very basic use of the platform’s (browser’s) features. Some who publish full on webapps make fairly sophisticated use of them. As a group they have more leverage over the OS/Browser than Lotus 1-2-3 and company ever had. That’s why IE can’t make crazy extensions W3C standards, just incremental ones.
- Compilers and languages have come a long way since C showed the way. The norm now is for runtimes to support multiple languages with a common API (see .NET, etc.). So no "single" language will need to be invented.
- NewSDK looks a lot like GWT. Sorry, Joel, Google isn’t snoozing. They may still get overtaken, but your narrative is a little off the mark.
What is the hardware, the OS, the GUI SDK and the Application? You could take the naive approach and say that Intel is still the hardware, Windows XP/Vista is the OS, but that doesn’t quite square with how I see things. I think that the browser is the OS, DOM and Javascript are the GUI foundation library, and Dojo, Ext JS, etc. are the RogueWave and ThinkC SDK’s of the Ajax age. As the OS/GUI library changes (new features/standards in the browser, etc.), so these SDK’s will have to scramble to keep up. They will be influential, but eventually more and more of core GUI functionality will make its way into the browser. The browser calls the tune, everyone else just dances along.
