Tuesday, January 25, 2005, 8:00:00 AM, you wrote:


> I am strongly opposed to PaceSimpleLanguageTagging.

I had two problems with xml:lang:

One was implementation difficulty - I think that the level of language
support in Atom seems a bit unbalanced. On one hand we can language
tag everything anywhere - which is a good thing, but unprecedented for
syndication formats; on the other hand we can only have a single title
and a single alt link. Perhaps the idea is that there should be
multiple feeds and language selection should be done outside Atom, eg
with Accept-Language, but this doesn't seem to fit well with "xml:lang
everywhere", as there isn't necessarily a language for a feed - only
for the components of it.

Anyway, I've heard enough arguments in favour of xml:lang that I'll
accept that it is a necessary cost.


My other issue was that we need to state which core elements are
language sensitive and which are language neutral, otherwise
implementors will guess and language tags will get lost. We also need
to state what happens to extensions. I'd suggest that if we go with
PaceExtensionConstruct, then Simple Extension constructs are language
sensitive (although I really don't mind them being language neutral
either - I'm not convinced either way). Structured Language constructs
MAY be language sensitive (so if you don't know how to interpret them
specifically, then you MUST preserve the language tag).


So I am now -0.5 to PaceSimpleLanguageTagging, but I think that there
are still some issues, so I think we need another Pace that makes it
clear what xml:lang applies to.


BTW:

When no xml:lang is in context, is the default language be
language neutral (ie: xml:lang=""), or is it determined by the
envelope, eg: a HTTP Content-Language header?


-- 
Dave

Reply via email to