I don’t think I have an account with the bugtracker, but it isn’t as public as this forum. I can easily post this here, and people can easily find it, so.
In the 1.5 PDF it explains that <technique><param> will validate the <param> element. But this is no longer true in 1.5. The PDF needs to be changed to caution readers somehow. In short, this is only true for <technique><COLLADA>, which most documents are unlikely to do (put a nested COLLADA “document” inside of a <technique>.)
To validate “<param>” in 1.5 it must be written <technique><param_type>. Mark Barnes says this was by design, but the point of this post, is the PDF says otherwise.
An alternative is to use the <technique> element’s xmlns attribute to change the no-colon-namespace to the COLLADA 1.4.1 namespace, where <param> etc. will be validated.
FYI: By “validate” I mean not that the document will be invalid if <param> is used. It just won’t be seen as belonging to the COLLADA 1.5 namespace, and so validators will ignore it. And COLLADA-DOM for example, will use the generic string-based model for the element, rather than the databinding model with binary representation.
I will try to write this PDF mis-documentation up for the bugtracker, but as I understand it, no one is fielding bugs, so it’s something I might neglect to do for some time.
EDITED: Also, to be clear. The <technique> element houses arbitrary user-defined data. Profiles use it to describe their “techniques.” The implication is, that the “technique” cannot reuse COLLADA’s elements, unless it knows the names from the schema, which are not part of the PDF documentation, or it declares itself to be using COLLADA 1.4.1 … which is admittedly a little off, since the context is a COLLADA 1.5 document.