Larkware
We get up early so that you don't have to.

Post-PDC Thoughts

By Mike Gunderloy
Saturday, November 01, 2003

It was a slow surfing day for me yesterday; I actually don't have anything bookmarked other than the Longhorn Architecture diagram. So, for a change, no Saturday Grind. Instead, a few ruminations on the stuff that was ballyhooed at this year's PDC.

I should preface this by noting that I wasn't there. This means that I've been free of the Microsoft reality-distortion field, but also that I'm relying on published stuff to form my opinions. Also, you should know that I think Windows user interface design has been in a steady downhill slide since roughly Windows 3.0. That said, and in no particular order:

  • Avalon - The new presentation subsystem. The important part here is the new Framework classes that give me control over the new user-interface widgets. The unimportant part is XAML. Why are people so excited about writing XAML by hand? Do they write resource files by hand? I'd rather write procedural code than declarative markup any day. Hopefully the designers will write all the XAML I ever need, because frankly, I am quite sick of typing angle brackets. Oh, and a prediction: Office 2006 will introduce user interface "innovations" that are (a) not part of the standard Aero user interface (b) impossible to do using the Avalon classes (c) annoying.
  • WinFS - The new metadata layer in the file system. No, it's not a new file system, it's stuffed on top of NTFS. Potentially interesting, but if past metadata-oriented tools have taught us anything, it's that metadata is only useful if tools exist to easily create it. Pop quiz: if you turn on the "Comments" column in the Details view in Windows Explorer, how many of your files have comments? Uh-huh. Well, searching is going to be really fast if everything boils down to SELECT WHERE NULL.
  • Indigo - "Unified programming model and communications infrastructure for developing connected systems". Yeah, I'm sure it's the greatest distributed framework ever. But you know what? I've had ways to build distributed applications since DCOM. I can count the number of times I've been paid to do so on the fingers of one hand. Distributed applications are sexy, but I suspect most developers never need this goop.
  • Visual Studio "Whidbey" - Looks like some useful innovations here, and this one will be out in a reasonable time (unlike Longhorn), so I'll be digging in soon. My main interest is in the new stuff in the FCL. I don't feel the need for any language innovation at this point.
  • SQL Server "Yukon" - I've been playing with this one for a while. It's nifty. It has so many new capabilities I'll never use them all. Most current DBAs are going to pitch a fit when they discover that all of their familiar tools have been replaced by a version of the VS .NET shell. But this is the no-brainer upgrade of the bunch for me.
  • Visual SourceSafe - The dog that didn't bark in the night. A few months ago there were hints on various MS weblogs about a great new replacement for VSS. Now it turns out we're just getting the same old file-system-based dog refurbished again. At this point, MS isn't a credible player in the source code control space. VSS is OK for small projects with a limited number of developers, but for anything beyond that, buy a real source code control system. At least my friends at SourceGear will be happy about this.
  • MSBuild - So, Microsoft is so paranoid about open source software that they've decided to reinvent Ant and NAnt with their own incompatible syntax. And it's more damned angle brackets. Is there a user interface for this stuff? If not, I'll stick with FinalBuilder.
  • Next Generation Secure Computing Base - I do not want these bits on my computer.
  • Longhorn Help - You realize, I assume, that it will end up being something like eight or ten years between help versions from Microsoft, now that they've killed off HTML Help 2.0. I expect I'll have everything moved over to RoboHelp's formats long before Microsoft gets back into the picture here.Also, is this stuff gonna work on other platforms, or only Longhorn? If the latter, it's not great for anyone trying to maximize sales.
  • ObjectSpaces - Microsoft gets into the object-relational mapping act two years after everyone else. I fully expect to see any innovations here copied in other products before MS manages to release.

Overall: lots of ambitious plans, but count on almost everything changing before it ships. Whidbey and Yukon are probably past the design-change stage, except in minor ways; if you want to concentrate your learning time on stuff with short-term applicability, start there.

Mike Gunderloy is the lead developer for Larkware and author of numerous books and articles on programming topics.

Home