Sorry its taken me a while to get back to you, RebelGeekz.
Tabular XML: Its just a medium, not a message and its has no meaning.
At least, that's what I took away as being your major thought here. Its pretty clear that if you're using XML purely as a transmission medium -- as it stands today -- its pretty easy to have a high meta-data to data ratio. And from that prospective, I completely agree. XML's virtuous verbosity is also its vital vexation. But that's just one use case for XML, and frankly, I don't think its a very good one: its just CSV++.
But what you've said did make me think a lot about what I initially advocated as a better choice: compressed-body XML. Over the last few days I've been reading a lot of arguments from both sides over Binary XML. I'm thinking I really don't agree with Mark Fussel that its a stillborn idea, but like other choices, I don't think it can really ever achieve the same level of ubiquity that pure XML has simply because its making more assumptions about the processing of the document. No matter how you look at doing compressed-body XML, somebody, someplace isn't going to be able to support it. And if you go down that path very far, XML as a medium quickly looses its appeal.
Another popular use case for XML is for messaging. In one sense, verbosity isn't so much of an issue for a message-based design because the payload per message really should be fairly small to begin with: you aren't transmitting 10,000 items of inventory data, rather, you simply transmitting the deltas of a few items assuming both endpoints are in the same relative state. I do think this is a good use for XML since we are frequently concerned about crossing boundaries blindly, so the less assumptions about the processing of a message that have to be made, the better.
Then there's whole concept that XML's metadata and positional cardinality has meaning beyond simple mapping. That's where I see the real value in using XML as a representational format since there's few alternatives that are as interoperable as XML is in this space.
So when it comes to adding the concept of a dynamic node-mapping alternative-delimitation schema like you've suggested here, I can see more value in it that I originally thought there was. It still strikes me as only part of the bigger picture though, and certainly one that may not be as easy to implement universally as we'd like. I do like it in the XML as the medium use case though.
As long as you that's all you use it for.