Mike Says

----

Hmm - I can imagine some cases where the document size would more than double using this scheme - not an unimportant consideration for high-performance worldwide document distribution.  Perhaps the ideal would have been to allow both encodings.

-Mike

-------

Again back to implementation dependent decisions.

Take a look at the previously mentioned article about EXI compression.

http://www.svgopen.org/2010/papers/3-Compressing_SVG_with_EXI/index.html

With EXI compression the current SVN model produces bigger documents then a node-based model.

So back to the "assumed performance" predicament.

Size is always an issue :)

Similarly when I first attempted to store documents into MarkLogic and query them they performed badly.

Turns out I had very normalized documents (basically a relational dump) ... and by de-normalizing the documents (which made them vastly bigger) performance improved 100x (sometimes more).   Who'd have guessed (I didn't!)

So ... where do we as designers make the decision about what are the important criteria ?