There's certainly a trade-off: when you provide a new construct you have to weigh the fact that it will be easier to write certain stylesheets against the fact that the spec will become longer, the implementation bigger, and the learning curve a bit longer.

I think the identity template pattern is so common, however - and it often increases the size of a stylesheet by 200% - that it's well worth making it easier to write. Our main motivation, though, was not to reduce the number of lines of code that users have to write, but rather to make it easier for optimizers to detect the pattern, which is particularly important (we decided) as an enabler for streaming transformations.