subState when joining segments

[Yves Savourel <ysavourel@enlaso.com>]

Hi all,

While trying to implement joining segments I ran into an issue not addressed in the processing requirements of the Segmentation
Modification section:

There is no guideline on how to handle subState if it has different values in the two original segments.

A first solution would be to take the subState value associated with the state value that is used in the resulting segment. So if we
have this:

<unit id="1">
 <segment id="s1" state="reviewed" subState="my:value1">
  <source>Sentence 1. </source>
  <target>Phrase 1. </target>
 </segment>
 <segment id="s2" state="translated" subState="my:value2">
  <source>Sentence 2.</source>
  <target>Phrase 2.</target>
 </segment>
</unit>

We would get this:

<unit id="1">
 <segment id="s1" state="translated" subState="my:value2">
  <source>Sentence 1. Sentence 2.</source>
  <target>Phrase 1. Phrase 2.</target>
 </segment>
</unit>

But that breaks down if we have no difference between state values and differences between subState values. For example with this:

<unit id="1">
 <segment id="s1" state="initial" subState="my:value1">
  <source>Sentence 1. </source>
 </segment>
 <segment id="s2" state="initial" subState="my:value2">
  <source>Sentence 2.</source>
 </segment>
</unit>

We would get this:

<unit id="1">
 <segment id="s1" state="initial"
  subState="???my:value1 or my:value2"???>
  <source>Sentence 1. Sentence 2.</source>
 </segment>
</unit>

One possible option would be to take the value that has the value string (i.e. the part after the prefix) with lowest rank in
alphabetical order. This would allow user to define their set of values accordingly.

Thoughts?
-yves