On Feb 1, 2004, at 6:57 PM, Michael Kay wrote:

>> In practice our schema fragment works great with forms that
>> have slots for firstname, middlename, lastname
>>
>
> Both my parents are generally known by their second Christian name (my
> father is English, my mother German, so this is not peculiar to one
> culture). Many IT systems - especially health systems - can't cope with
> this, and regularly address them by their first Christian name, which 
> is
> not only impolite, it can create the risk that they don't get the right
> medication. Before you assert that your schema fragment "works great",
> you need to ask the victims of your IT system whether they agree.

Of course you are knowledgeable enough to understand the difference 
between an IT system and a schema. I have no IT system, and hence no 
such victims. I certainly won't accept responsibility for the world's 
healthcare IT software (most of which is horrible).

But indeed I understand the issue, which is why one may flag various 
name components with attributes ... something that is rather more 
difficult to do when name components are tokenizen in a string.

>
> (And when I say "Christian name" rather than "given name", I mean it.
> You might call it a given name, but they don't.)

That's the tag name, and despite it being related to an English 
language term, you are well aware that there is not necessarily any 
semantic relationship between the tag name and the semantics of the 
content. For all we care a name could be represented as:

<foobar>
	<fribbut>John</fribbet>
	<dongle>Smith</dongle>
</foobar>

and as long as the software knows how to present it to the person, all 
is fine ... indeed it is with the software that the semantics resides.

>
> I had a colleague who was generally known by his third Christian name
> (A. P. Graham Brown). Most US-based IT systems (and government forms)
> fall in a heap on that one.
>
> Under UK Data Protection legislation, if a data subject tells you that
> your data on them is inaccurate, you are obliged to correct it. If your
> data model can't represent the information accurately, that's no 
> excuse.
>
Data model? I am talking about XML. The XML certainly can represent the 
information. I generally call people by the names they give me, but 
frankly the name a person has is meaningless to the actual surgery that 
I do. In any case I've not heard actual issues with medical records as 
a result of the XML. If you are having an issue in the U.K. it is 
highly likely at least partially a result of not using XML at all, 
rather than a result of the XML (regardless of which schema you are 
using).

In particular to the above, the problem is with the *form* not the XML. 
Indeed:

<person.name>
	<given>A.</given>
	<given>P.</given>
	<given type="familiar">Graham</given>
	<family>Brown</family>
</person.name>

The goal of this schema is not to constrain the way software might be 
used to solve such issues, rather to be suitable as a serialization 
format for such software (e.g. good and bad software).

Jonathan