[osis-core] Schema: type on language

[Chris Little]

Todd,


Todd Tillinghast wrote:

> Chris,
> 
> Thanks that does make it much clearer.
> 
> How do we interpret <language> with no "type" attribute?
> 
> Should "type" be required or defaulted?

I don't think type is necessarily required for <language>.  If we like, 
we could make "base" the default value, since it is most generic and 
most common.  We could require it in prose for > level 0 conformance. 
(As with xml:lang on <osisText>, I can live with it if others feel this 
should be required.)


> Does this eliminate the "x-SIL-ALS" form or is that just for xml:lang?

relevent examples:

 >>(Albanian)
 >><language type="ISO-639-1">sq</language>
 >><language type="ISO-639-2-T">sqi</language>
 >><language type="ISO-639-2-B">alb</language>
 >><language type="SIL">ALS</language>
 >><language type="IETF">sq</language>

I would recommend that we identify a canonical order of precedence if 
RFC 3066 doesn't, namely: ISO-639-1, ISO-639-2-T*, IANA, SIL, LINGUIST. 
  That is, you should only use an SIL code if none exists in either the 
ISO or the IANA code lists.

So, while ALS is still a valid SIL code, the IETF form should be 
identical to the ISO-639-1 form, since it exists.

(* I chose ISO-639-2-T rather than -B because it is based on ISO-639-1 
whereas -B is based on MARC language codes.)

> How do the xml:lang and <language> coordinate?

To expand on my other reply....  xml:lang on <osisText> should match 
<language type="base"> (and/or in some cases, <language 
type="translation">.  xml:lang on <foreign> and <q> should probably 
match ith <language type="quotation">.  The others are probably a little 
more complex to divine.  But if a language is identified in an xml:lang 
value, there should be some corresponding <language> element in the 
header, and if its type="IETF", they should match.  (Should we require 
in prose that a <language type="IETF"> element occurs for each language 
in order to match with xml:lang values?)

Interlinears might be a little different.  If the interlinear is done by 
using <w gloss="">, there's not necessarily going to be any indication 
of the interlinear's language except in the <language> element itself.

--Chris