Sam Ruby ...how are decimals and dates to be handled?
There's the rub. While a text format data isn't optimal, it is, more or less, equally non-optimal on every platform. Just about any binary representation you choose for a particular data type gives some platform an advantage at the cost of another. Would you like big-endian or little-endin with that integer?
Is a single standard binary serialization doomed for this reason? Peter, how widely supported does it need to be a good thing? Do we count computers or platforms?
What about a format that keeps text for all content but replaces element and attribute names with numbers after the first occurrence? Is it worth it?