From y.umaoka at gmail.com  Tue May 26 14:55:09 2009
From: y.umaoka at gmail.com (Yoshito Umaoka)
Date: Tue, 26 May 2009 17:55:09 -0400
Subject: [loc-en-dev] Definition of "candidate locales"
Message-ID: <4A1C653D.40708@gmail.com>

The API doc for ResourceBundle.Control#getCandidateLocales says -

--------
The sequence of the candidate locales also corresponds to the runtime 
resource lookup path (also known as the /parent chain/), if the 
corresponding resource bundles for the candidate locales exist and their 
parents are not defined by loaded resource bundles themselves. The last 
element of the list must be a root locale 
<http://java.sun.com/javase/6/docs/api/java/util/Locale.html#ROOT> if it 
is desired to have the base bundle as the terminal of the parent chain.
--------

According to the agreement in the former discussions, we are going to 
introduce the special lookup path for supporting backward compatible 
locales.  For example, for a given locale "he_IL", the lookup path will 
be "he_IL", "iw_IL", "he", "iw".  The current ResourceBundle 
implementation assumes that the lookup path is equivalent to the 
ResourceBundle's parent chain.  Therefore, when both he bundle and iw 
budnle exist, ResourceBundle#getObject go through bundle he, then iw to 
locate a resource.  I'm wondering if we are OK with the behavior.  
Another possible option is to separate the parent chain from the lookup 
path - for this specific case, if "iw_IL" exist, the parent chain will 
be "iw_IL", "iw" and ROOT.  But if we want to go for this option, we 
probably have to make this special case only for JDK's own Control 
implementation.  Anyway, I would like to raise this topic for discussion.

-Yoshito