Not to mention it is ignorant of prior art, by which it is inside-out. Why would you embed the data in an XML document inside comments, when you could build a mime/multipart message of which the XML document is the principal part? The good thing is, that approach respects all the existing infrastructure. Rather than educating the XML parser about strange-looking comments, you only need to educate the existing extra transport layer underneath the parser, which in the case of clients such as browsers already has most of the plumbing in place.
]]>