In regards to why one would NOT want to use this approach with dynamic web sites, in the time I have written this, I do have a few scenarios where using something like ASP.NET controls has been useful (despite another blog post compliaining about how leaky its abstraction is!).

That is, I have found in an application that requires extremely rich desktop controls and interactions beyond rich web applications, with an XSLT type of framework, we end up trying to recreate all the view state and other mechanisms already built into ASP.NET. In that scenario, a trade-off may be applicable depending on where the majority of effort for development/maintenance will likely fall.

I am not completely convinced of my own thought here, and one day will try to blog this aspect and explore it further.

Thanks for your comments. I agree with you — web frameworks are a good idea but would be nice if their template languages could be improved and made more consistent.

I think I can understand why some people in both closed and open source shy away from it. There’s always the other angle that if your framework is in say C#, PHP, Python etc then a templating language based on that is also nice.

With ASP.NET for example, for very complex and form-heavy applications (as opposed to web sites) sometimes their control model with view state is actually okay (that being said, other times I do feel XSLT can help remove some bloat).

On the unit tests, I apologize for not posting something sooner (this post was written in Feb, and I still haven’t got round to it!) Although there are a number of ways, the technique I have found useful is to use code like PHP, Java, or C# etc to run a unit test that generates output from an XSLT and do an XML diff with an expected output file. Will try to post something soon when I can get some time

Look forward to your blog post reply!

This is a fantastic post, thank you. I am working on a reply on my own blog page, but in brief it is heartening to me to know that there are other developers in the world using and enjoying XSLT as much as we have over the past 8 years or so. Our entire SaaS CMS is based around XML and XSLT, and we are still finding new ways to extend our templates for our customers.

Your stories of replicating sites that use XSLT templates strike a chord with me, as we have saved our clients a lot of time and money when upgrading, extending or duplicating web sites thanks to XSLT.

Web Application Frameworks are such a great idea, and wonderful from a web developer’s standpoint, but it is often disheartening to read a framework’s documentation and find yet another proprietary, non-standard template language for rendering HTML. It seems ironic to me that in a world of open-source, open standards are not always embraced.

I would like to see all frameworks introduce support for XSLT so that template designers are spared from learning yet another template language.

Anyway, great work and I’ll be sure to post a reply to your article in a couple of days.

- Daniel

PS. I would be really interested in how you have done unit tests in XSL, and would appreciate a quick post, even if just in a comment.

The XSLT being done in the browser with JavaScript is interesting. I had tinkered with this some time back and found a few inconsistencies and limited support. Not looked at it for a while now.

How have you found it, or do you have a blog post on it?

Thanks much for the extremely well-thought out and informative article!

@Christophe: Thanks for the links. The Stylesheet Schemas is interesting, for if I understand correctly, the implication is that you could use this to test if your XSLT needs updating because code elsewhere has been refactored? (assuming you can keep the stylesheet schema updated without much pain when refactoring other code that transforms to XML, etc…?)

Regarding errors etcetera: are you aware of Bob DuCharme’s article “Using Stylesheet Schemas” on XML.com? http://www.xml.com/pub/a/2005/04/06/tr-xml.html
Of course, this helps only against one type of errors.