the case against .NET

I’m aiming to change the header graphic regularly. The opening image shows the River Thames in London, taken from the new Hungerford Bridge on a dull afternoon, Saturday January 29th 2005.

I was in London for the visit of Sheffield United - the football (soccer) team I support - to West Ham United, for an FA Cup 4th round match. The game was drawn 1-1, with Sheffield United later winning the replay on penalties.

(Photo taken with Casio Exilim EX-Z30.)

In my opening post I mentioned that, in the past, I might have called this blog “the case against C++”. In other words, I’m taking issue with developers who unquestioningly use languages and technologies without looking at what’s most appropriate for their application (i.e. the herd mentality). .NET is simply the most recent cause/object of this lemming-like behaviour.

Richard Hale Shaw is running a series of posts looking at moving from C++ to C#. He begins with some good stuff on why C++ is such a bad idea. Well worth a read!

Nick Bradbury reports on the creation of AttentionTrust.org, a not-for-profit “dedicated to promoting the basic rights of attention owners”.

As Nick suggests, the concept of attention data seems pretty woolly to most of us. The idea is that, as society becomes increasingly web-oriented, more and more companies are gathering information about things we’re all interested in. The information held about you is, increasingly, an externally-held defintion of your self. As such, this information belongs (or should belong) exclusively to you.

AttentionTrust.org is campaigning to ensure that each of us has full control over our own attention data. Nick is surely right when he says that this issue is going to become far more important as time progresses. I encourage you to support the AttentionTrust campaign.

First off, I’m a big fan of Robert Scoble. His blog has a high signal-to-noise ratio, and he’s generally very readable. Above all, the fact that Microsoft allows/encourages him to blog has been largely responsible for the much more positive perception of Microsoft now held by many people.

However, someone needs to take issue with Scoble’s professed status as an amateur blogger. In his response to a rather pointless article in The Register about IE7 beta 1 and third-party plugins, Scoble reiterates that he doesn’t get paid for journalism.

I’m not sure whether he’s saying that his blog isn’t journalism, or that Microsoft doesn’t pay him to maintain it. Either way, it’s clear that he posts during ‘normal working hours’, and that Microsoft allows him to do do (and thereby encourages the practice). It’s also pretty clear that blogging is a form of journalism.

In other words, Scoble should be prepared to acknowledge his blog’s status as a (paid for) element of Microsoft’s marketing strategy. Or have I missed something?

Dave Winer reports on a couple of patent applications from Google and Amazon.

The Google application is particularly bizarre. It covers the inclusion of targeted ads in syndicated feeds (e.g. RSS). Even if Google is granted this patent, it’s impossible to see how they could defend it while maintaining their “not evil” image.

It’s sad to see these great web innovators falling back on patents in this way. Like an athlete who looks over his shoulder, they’re sending out a message that they can no longer rely on continuing creativity to stay ahead of the pack.

Scoble has been asking whether venture capitalists (VCs) are investing in .NET developments.

This is a bizarre, geeky kind of question. A good VC won’t care whether the project rests on .NET or any other platform. He’ll be looking for innovation leading to return on investment.

So where does .NET offer the potential of productivity gains? One place only: the server. Here, there is no need to burden users with 20-something MB of extra downloads/installs, and concerns about memory and performance overheads. In any case, .NET offers client developers minimal productivity gains over best-in-class Win32 tools.

On the server, though, .NET can shine. There’s real innovation in ASP.NET. Combine this with IIS 6 running on Windows 2003 and you have a compelling server platform.

Around 10 years ago, Microsoft made a costly mistake in pushing ActiveX Controls for use in web pages. Then, the main competitor was Java applets.

Ten years on, Microsoft is still fighting a phantom Java enemy by pushing .NET on Windows client computers. The truth is that Win32 already had this market sown up. In undermining Win32, Microsoft risks losing the initiative in client computing altogether.

I had my worst ever software installation experience a couple of days ago. I’d previously thought that Visual Studio .NET installation was pretty bad, simply because it’s so painfully slow. However, the VS.NET setup process is a model of excellence when set alongside that for Borland Delphi 2005.

In mitigation, I should say that Delphi 2005 is a very feature-rich product, incorporating Delphi compilers for Win32 and .NET, as well as C# language support. However, the installation experience would make a textbook example of how not to handle such things.

On opening the box (Delphi 2005 Professional), I found a total of 4 CDs. At least two are required for a typical install. This would not be too big a problem, were it not for the patching process (see below).

Nevertheless, the first problem: bad choice of media. No DVD-ROM was provided, and no DVD option was available when buying.

The patching process is at the root of most of the problems with Delphi 2005 installation. You see, there are three patches: Update 1, Update 2 and Update 3. Each of the updates must be applied in the correct sequence.

So, the second problem: no cumulative patches are available.

Having installed Delphi 2005 from the original media, including swapping CD2 for CD1 when directed, I started running Update 1. I soon found I needed to insert first CD1, and later CD2, in order to install Update 1.

This is the third problem: the installation database is not copied to the hard disk, so the original media are needed each time a patch is installed.

That’s right - I would also need to insert both CD1 and CD2 in turn when installing Update 2 and Update 3. That makes a total of 8 CD swaps to complete product installation. Had a DVD been provided, I could have left the same disk in the drive throughout.

At the end of Update 1, I’d spent around 40 minutes sitting by my PC waiting for it to ask for another CD to be inserted.

I then moved on to Update 2. Lo and behold, this failed to run at all. It threw up some message about the base software not being at the right version. I double- and triple-checked that I was installing the updates in the correct sequence - I was. I then spent around 20 minutes with Google, to find that actually there were two versions of Update 2. The first version, the one I had, would fail to update certain copies of Delphi 2005.

The fourth problem: inadequate quality control. I appreciate that such things happen in software development, and I could have lived with this had it not been compounded by the other issues covered here.

So I downloaded and installed the corrected version of Update 2, and then ran Update 3 (lots more CD swapping). At the conclusion of all of this, I’d spent around 2 hours installing Delphi 2005.

Only later did I realise the irony of the fact that Borland’s current marketing campaign is centred around “Software Delivery Optimization”. Maybe that should be “Software Delivery Obfuscation”? Don’t call us, Borland.

So Longhorn is to be “Windows Vista”.

Actually, I think this has a nice sound to it. My dictionary, however, says a vista is “a long narrow view”. I think many would agree that Microsoft’s biggest current problem is its narrowness of vision.

Paul Gustavson makes the point that users aren’t asking for .NET apps. If they want a Windows program then they’re quite happy with the Win32 API. If they want cross-platform then Java dominates.

The key point is that the Win32 API remains Microsoft’s unique selling point. It’s easily the most used API, and it’s not about to go anywhere.

.NET falls uncomfortably between two stools. It wants to supersede the Win32 API, but Microsoft has too much investment in Win32 to allow that to happen. And it wants to be set free to be cross-platform, but Microsoft won’t take that risk either.

The message to developers is clear: you’re more competitive on Windows if you stick with Win32, preferably with a grown up language and class library like Delphi’s Object Pascal and the VCL.

A neighbour found this bird trapped in his garage this evening. He carried it to a nearby tree, where it ate hungrily and quickly regained strength.

This is the first goldcrest I’ve seen in local gardens in 8 years living in this area.

(Photo taken with Casio Exilim EX-Z750, default settings.)