[xliff-tools] Approved attribute

[Jean-Christophe Helary]

On 5 avr. 07, at 10:59, Bruno Haible wrote:

> Rodolfo M. Raya wrote:
>> You can't use an <alt-trans> element
>> because you don't know what to put in the <source> element.
>>
>> In an <alt-trans> element the content of the <target> element must  
>> be a
>> translation of the corresponding <source>. The original source of a
>> translation marked as fuzzy in a PO file is unknown.
>
> But in <alt-trans> the <source> can be omitted, and the 'origin'  
> attribute
> is not required either.

It is counter intuitive to put an "alternative translation" when the  
source is different.

In the case you describe, it is do to PO limitations and such  
limitations should not affect the design of a XLIFF file.

and <alt-trans> that does not have its own <source> is expected to  
share the source with the segment is it made an alternative  
translation of.

If the source is different it is expected to be indicated.

A practical reason for that is when one extracts data to create a  
TMX: it does not mean anything to associate a translation to a source  
of which it is not the translation.

> Also, <alt-trans> seems better suited because there can be several  
> of them
> in a <trans-unit>: When an approximate translation is searched for  
> in several
> translation memory databases, the result will be several  
> "alternate" or
> "proposed" <target>s. The PO file can store only one approximate  
> search
> result per <source> - this is the role of the #,fuzzy flag - but  
> when looking
> at the XLIFF format in general, the way to store several approximate
> matches from translation memories are several <alt-trans>, no?

At the condition that the source is informed. The alt-trans  
description makes it seems obvious that when the source differ it is  
indicated with the possibility to add the optional match-quality  
attribute.


Jean-Christophe Helary