Comment by chinpokomon

I've seen XSLT done well, and I've seen the mess it can be when written by someone with only a passing knowledge. As a consultant I usually have dealt with the latter, and that usually goes hand in hand with a poorly designed XML schema.

One space that XSLT can demonstrate its strength is when transforming some horribly serialized interoperability data structure. If the system from which you are receiving data, produces terse XML, you aren't going to solve anything by rewriting the upstream system to produce equally lousy JSON. If you don't have the ability to fix the upstream service to produce better structured data, XSLT and XPath are wonderful tools to morph it into something more manageable. That transformation process is better written with XSLTs than it is in trying to do the same thing by slurping the data directly into some business object and trying then to work with a bad model. Don't go down the path of "garbage out, garbage in."

If you have access to both sides of the process, it might be worth rewriting the upstream system, but when working with a legacy system XSLT might be the best glue technology in your arsenal.