I am not sure I agree with you on silverlight looking at corporate intranet apps. Nothing I have seen on silverlight has ever said that. If you look at the silverlight showcase site (http://silverlight.net/showcase/), none of the apps look like dashboard /CRM/ etc apps. Also as far as I know, Silverlight has no advanced components to make creating such apps easy (like datagrids, tree controls. charting controls etc). They may appear as additional products you can buy later but they arent part of the core product.

I think silverlight is closer to Flash than Flex. But if the runtime gathers enough steam, a framework for silverlight will appear soon enough.

In terms of product decisions, I think you are assuming that all those apps will run outside the enterprise’s firewall in which case the 90% penetration is justified as a critical point.

However, I’d bet a large sum of money that each enterprise has a lot more more internal corporate apps than those deployed on the www. And these corporate companies are who Adobe + M$FT + everyone else is competing for. I’ve seen so many internal apps run on Weblogic, Websphere, or any other >10k per cpu server in my time as I’m sure many of us have….and in this case, a 90% penetration rate doesn’t carry that much weight when they / the corporation controls the config on each desktop.

I don’t disagree that flex outputs a nice product, but going to the silverlight site, for a business ria app, I don’t see a distinct difference in the output – in which case labor, and time to build will obviously be an important factor.

I’ve built flex apps myself, and while it’s quick to create simple pages, i’ve found complicated pages to be much more difficult + I’ve found mundane tasks like code formatting, organizing imports, refactoring etc. to be frustrating and a waste of time b/c of a poor tool. I’m just assuming mind you b/c I haven’t done silverlight dev myself — but M$FT has always put out good tools to really RAD their applications(It helps to just build tools that work on 1 operating system!) starting w/ the original VB(although they probably get a mulligan for Visual C++)

Clarification – MS Visual Studio .net is the most widely used IDE in the world. By a long mile actually…and again, I’m not a .net developer…just stating facts.

Adobe should look to Jetbrains for tool inspiration — they make the best stuff bar none.

I am sure MS’s tools for silverlight will be amazing, they do have a bigger team than Adobe on tools and Visual Studio is an awesome IDE. But arent product decisions about the final application coming in from the business folks and unless the silverlight plugin reaches a 90% penetration, trying to convince business to use silverlight for anything besides prototypes would be hard.

I hope MS does a kickass job on the tools. Adobe needs the competition. Flex Builder 2 was a first stab but FB3 seems to have a lot more support for refactoring and debugging.
I agree that the unit testing framework for Flex needs a lot to be desired.

On the other hand Adobe has an army of designers who could really enhance the user experience on RIAs and MS’s lacks that. Imagine improved workflows from photoshop to Flex and how much that could reinforce Flex’s current dominance.

Secondly, I expect MSFT to build decent(definitely better than FlexBuilder) tools for silverlight. It’s not like Adobe starting an IDE from scratch, they have the knowledgebase / codebase already and know how to do it well.

I expect that .net shops will develop in silverlight, and other shops will choose from Flex or web 2.0 / html + javascript

Also, custom component creation is in fact harder than it is in java swing. The actual code you need to write is the same, but the fact that Flex Builder isn’t nearly as good as the standard java IDE’s means little shortcuts for moving up and down the inheritance chain, overriding methods from your parent etc are not there in FlexBuilder. If you open up one of Ely Greenfield’s custom components and try to work out what is what, you’ll see what I mean — FlexBuilder gives you no hints on what methods the child overrode etc.

Btw if you are an agile shop, I’d think twice before choosing flex. Flex unit is a ‘not good enough’ first attempt to test their actionscript code, but not the mxml – which will be the main view of your ui. The only way you can really test the gui is by buying some mercury product. Appalling really why Adobe only conspired w/ Mercury and didn’t assist the ‘open source’ world in any way to test the gui.

@Robert:
Flex and Silverlight is going to be interesting if silverlight can get the plugin base high enough. Whats a little weird to me is Silverlight seems to be targeting the developer community (especially with the language independence) but the applications they showcase are more what a web developer would be called to do rather than a software engineer. I am not sure how many companies will ask their C# developers to create an application with tweens, fades and other eye candy. If the battle doesnt spill to those domains and Silverlight just remains something in a web developer’s arsenal. then its future would seem more rocky.

AS3 is also a standardized language. Web developers who already know Javascript and PHP (as opposed to so-called “real languages”) can transition into Flex relatively easily. Libraries are relatively easy to find already. If you prefer a language with stricter typing rules, the haXe open-source project is available; it has similar syntax to AS3, but adds an implicit typing system.