If you are so bandwidth constrained that deflated XML won't do, then I doubt deflated JSON would be good enough either (and that exists anyway, Matrix).
However, on your latter point, I am in full agreement.
My position is that compression ratio is a largely meaningless metric in this instance, so using it as a method for justifying the use of XML (as inferred from "XML compresses really well") is also meaningless. That does not translate to "XML is bad for XMPP", I actually think it is fine, it means "XML compresses really well" doesn't add anything much in the way of justification.
I've spent way too much time on this thread already, so take what you will from it, it is all you from here on out, I am done here.
The only thing I was saying was exactly what I said, within the context of XMPP, XML compresses really well - I was reinforcing the parent comment. How that relates to metrics in other contexts wasn't what I was talking about at all. I generally avoid XML nowadays, and so your argument with me isn't something I would dedicate thought to in the first place. I see no point in taking anything from this nonsense.