Stunted imaginations

A reader pointed me to this post by Jon Udell in InfoWorld. He suggests a native data type for XML in the JVM and the CLR:

Whenever I see the emergence of per-programming-language variations on a theme, I wonder what can be abstracted. In this case, I wonder about the feasibility -- and the desirability -- of pushing the notion of a native XML datatype down into the JVM and CLR/Mono. In theory, the benefits would be:

• A single robust implementation
• Smoother transfer of experience across programming-language domains
• A common focus for storage implementations

This is the clear trend that the static languages (the popular ones) seem to take - whenever an idea pops up, lard some additional complexity onto the language. This is in fairly stark contrast to the Smalltalk way - we had libraries, not complexity. Why is that better? Well, say someone goes ahead and follows Jon's suggestion. Now developers are forever wedged into whatever notions that developer had (think primitive data types, for instance). In a properly constructed library (i.e., one without any final classes), follow-on developers can customize behavior to suit the needs of their projects. With this kind of hard baking-in, developers instead end up working around limitations forever