[tei-council] <content> vs <mixedContent>

[Lou Burnard]

On 06/10/14 10:40, Sebastian Rahtz wrote:
>
>   - Do we really know what we mean by a "text node" anyway?
> Indeed. The assumption is that it maps to a W3C “string”. But this
> a) doesn’t map to the common pattern we use of “text | <g>”, and b)
> gives no scope for other data types (e.g. token, number, date etc).
> So if we have to provide a way of pointing to other datatypes, it
> may well be a bit redundant.

The proposal was to use the element <textNode> as a stand-in for those 
parts of a content model where we want to permit a text node. As such it 
is really a macro, syntactic sugar for <macroRef key="something"/>
Since we are not entirely sure what "something" should be, and since we 
are quite sure that whichever we decide people will occasionally want 
something else, there is little advantage to special casing one 
particular kind of text node.
>   Special-casing it as a shortcut implies
> its the norm, when I suggest that it isn’t (“text | <gi>” is our norm).
I think by "<gi>" mlearned friend means "<g>" . Note also that this is a 
"norm" in  TEI practice, not in XML schemas in general. pureODD is 
intended for use outside the TEI too.

> Lou, I think, still believes that it maps to many people’s common-sense
> view of a schema, and so should be kept.

No, my view on <textNode> is now as expounded above.
>
> Re <mixedContent>, what’s missing is the definition of what “textual
> nodes” are. I feel the same way as about the “text node” in <textNode>,
> that mapping it to “string” implies that this a good norm which is correct for most
> cases.  Lou and I see this differently. YMMV.

It is true that we need to decide what the "text" part of "mixedContent" 
is. However, unlike textNode, the user can easily change the meaning of 
<mixedContent> by redefining its content model.

Note also exactly the same uncertainties about what "text" means (for 
example, are spaces normalised or not?) apply to the @allowText proposal.