The role of country codes.

[Jon Hanna]

> > I have failed
> > to convince that the primary differences between en-US and en-IE aren't
> > spelling,
>
> I'm convinced.  I'm just not sure what a computer can do about such
> differences as exist.  Speech classification, while important in theory,
> is still a very marginal use of RFC 3066 tags.

If you indulge some pseudo-code:

DocX.version(0).lang='en-US'
DocX.version(1).lang='en-IE'

UserX.lang='en-US'

DocX.getVersionForUser(UserX)

Code like that represented by the above is hardly wonderfully sophisticated.
It is hardly rare either however. The system doesn't need to know the
details of the differences between the two dialects, just that differences
exist.

When atomicity can remain throughout (as it can in this case since one of
the available languages matches the requested) this is fine no matter what
system is used. When atomicity cannot remain the semantics (however weak) of
the tagging system become more important.

When you add a degree of sophistication (it's reasonably easy to convert the
majority of Old World English spellings to New World English spellings
programmatically) then you might be able to adapt much better for one part
of the mismatch than another.

KM is increasingly using documents which contain spoken language as part or
all of their lingual content. In such contexts script is irrelevant.
Conversely many systems can happily do something useful with any text of a
particular script agnostic to language (coding something that can render
English, new orthography Irish and French without knowing which of those it
has is pretty easy). The obvious hierarchy is obvious only in a given
context.