As I noted in Bug 9965, Canonical XML - aka C14N - have some requirements that looks like relevant design principles for Polyglot Markup - at least in some stake holders' opinions. For example C14n requires:

* that <li/> should be written as <li></li>
* that document uses UTF-8
* that class="   one     two" is written class="one two" 
* XML declaration must be removed

For overview: http://www.w3.org/TR/xml-c14n#Terminology

Hence i suggest as a design principle for the Polyglot Markup that the XML requirements should be based on Canoncical XML - except when that principle is incompatible with HTML5. (One place where it is incompatible is in C14N's requirement to remove the DOCTYPE).)

Positive: Adopting such a principle would make it simpler to justify and agree about the design of Polyglot Markup. 

Negative: The requiremetns of C14N are, however, stricter than the the _real_ requirements of Polyglot Markup.

JUSTIFICATION: It seems to be the case that quite a few people with a stake in Polyglot Markup wants to base the polyglot requirements on something that is stricter thant the the _real_ requirements. And because of that, I think it is important to have something solid to base those extra requirements on. Rather than randomly picking one's favorite restrictions/permissions.

(For example, there are conditionas when it is not necessary to write <li/> as <li></li> - even when the document is meant to be polyglot. But instead of defining the exceptions  - or our own restrictions - we could simply pick Canonical XML's rulings as base.)