=?utf-8?Q?Re=3A_=5Bxml-dev=5D_The_semantics_of_an_XML_document_is?==?utf

[Michael Kay <mike@saxonica.com>]

> 
> To me, the problem of semantics is not schematic or functional, but historic: finding out what on earth people have actually stuffed into the element and then figuring out which ones of those ideal hoped-for functions don't produce garbage, and what to do about it. 
> 
> So semantics is about what you have, not what you wanted (schemas), nor what you may need now (functions).
> 

Which brings us on to tag abuse...

I remember we (ICL) developed a system for a news agency to transmit news stories to its client newspapers. They were tagged with the kind of story (sport, business, politics, etc). Having a handy mechanism for communicating electronically with their customers (this was the 1980s), they invented a new kind of story for sending invoices to their customers, and one day, inevitably, one of these invoices got published on the business pages of a national paper.

They had designed the invoice document type, of course, to pass all validation checks so they could get it through the system. Schemas are not there to ensure the information is correct, they are there to make users jump through hoops.

Michael Kay
Saxonica