OASIS XML Localisation Interchange File Format (XLIFF) TC

RE: [xliff] Amended XLIFF Teleconference Minutes - 11 Feb '03

  • 1.  RE: [xliff] Amended XLIFF Teleconference Minutes - 11 Feb '03

    Posted 02-13-2003 17:02
    Outcome
    Doug suggested that we include a section ‘helpful tips on how to migrate from 1.0 to 1.1’
    IT will be stated in this section that in order to have xliff 1.0 documents validated against the xliff 1.1 schema, the 1.1 namespace will have to be declared.

     

     

    3.2 Co-ords for <bin-unit> ?  (item 15 in tracking report)

     

    Summary

     

    John Reid discovered a need to specify co-ords of an icon for example in a dialog.  However, this is a <bin-unit> and does not have co-ord attributes. As per his mail, John would like to see coord, size-unit, maxheight, minheight, maxwidth and minwidth as attribs of <bin-unit>.  Where these items could then be controlled using the reformat mechanism.

     

     

    Discussion

    Mark is worried that this assumes that the bin-unit is being assumed to be a graphic in a dialog.  Also how should other binary units be handled, say the bin-unit is a wave file, it may have its own attributes that need editing, but these are not supported.

    John agreed that the need for coord, etc. are specific to the instance of a graphic on some kind of form.

    Mark is against refining the <bin-unit> element at this point.

     

    John  suggested that an alternative would be to allow a bin-unit in the trans-unit.  This would allow graphics to have co-ord info.

    Mark queried  - What about numerous translation information without having to specify the binary graphic each time?

     

    Yves handles this situation by putting the bin-unit somewhere in the doc and then a trans-unit references those from somewhere else in the doc.

    Yves said the tool would know what to do and how to find the binary objects. 

    John says this is not very interoperable.

     

    Mat suggests that we could link bin-units and trans-units by giving them the same id

    Yves - the NMTOKEN definition of the 1.0 schema would allow this.

     

    Yves, John,

    The main problem is that we need a mechanism within the spec for referencing another unit, i.e. reference a <bin-unit> from a <trans-unit>

     

    Outcome

     

    John will formally write and suggest how we could allow referencing in this way.  There may be a tu-ref and a bu-ref attribute that allow trans-units and bin-units to linked through their ids.

     

     

     

     

    3.3 Attribute Value Naming Convention (item 16 in tracking report)

     

    Summary

     

    Yves pointed out in his mail that at last weeks meeting it was suggested that the guidelines for using lowercase in the attribute values be removed.  Because the spec currently states ‘Attribute values are case sensitive. It is strongly recommended that lower-case values are used. The specification recommends a number of values for some attributes, these are all lower-case.’, we probably need to vote on any change to the guidelines.

     

     

    Discussion

    The discussion broadened into a number of points

    Attribute names – what case?

    Attribute values – what case?

    Attribute values – allow white space?

     

    John suggested that we remove the word ‘strongly’ from the spec.

    Mark suggested we have no whitespaces in attribute values, but allow CamelCasing.  A space for example, as per his previous proposal, is considered a delimiter in attribute values.

    It was pointed out that we must also abide by our guideline and that it should also apply to the x-somename attributes.

    Tony - would like to get rid of whitespace in attribute names and values, otherwise there is confusion.  Not worried about capitalisation.  Shouldn't have different naming conventions for enumerated lists and extended lists.

    Christian queried whether we have an issue for backward compatibility with the new naming convention

     

    John pointed out that where an attribute’s content definition in spec is defined as ‘text’ - 1.0 docs may very well have whitespace.

     

    Outcome

    1.  Keep lower case for the attribute names - no objections

    2.  Remove the word 'strongly' from spec  - because where attribute definition is 'text', we can't verify – no objections

    3.  Allow whitespace for attribute values - because if we change this then 1.0 documents containing white space in attribute values will fail 1.1 validation

     

     

    Aside

    Mark pointed out, as a related aside, that all suggestions for attribute values so far in the spec doc need to be made lower case.

    These are still under discussion and need to be agreed yet.

     

     

     

    3.4 Extension Points (item 17 in tracking report)

     

    Summary

    From Yves' mail.

    Where we have extension points, they are defined as ‘##other’.  Another option here is to define them as ‘##any’.  The difference is described here.

     

    Any - allows xliff elements to be included here

    Other - allows any elements except elements from the current namespace ( xliff in our case)

    'elements from any namespace where that namespace is not the current namespace'

     

    The issue is that if we use ‘any’, then users can conceivably embed xliff constructs at any extension point within another xliff document.  I.e. embed xliff within xliff.

     

    Discussion

     

    So, the discussion centres on whether users can add xliff elements at our extension points, though that is not what we intended these elements to be used for.

     

    Allowing this Any is obviously more flexible for the user.

     

    Christian would not like users to have to define totally new structures very similar to already existing xliff structures just because we don’t allow them to embed the already existing ones (be specifying ‘other’)

     

    Enda argued that from a tools perspective allowing by specifying ‘any’ and allowing xliff constructs to be embedded at points in an xliff doc where they were not intended would result in very messy and difficult to read xliff documents.

     

    Yves pointed out that it would be a way for us to refine xliff by seeing where our spec is limited, Enda, said this is a major issue for the tools.  Tools won't know what to do with, say a <trans-unit>, that is found in a place it was never expected.  We have already agreed that tools should ignore constructs they don’t understand.  This muddies the waters a bit because, in theory, we do understand <trans-unit>s.  However, if we come across a <trans-unit> as a child of a <trans-unit>, what do we do?

     

    Tony suggested that perhaps it would be obvious that these are extensions and not part of the core document because they will be prefixed by a namespace eg. urn:xliff.trans-unit.  Research and samples are needed on this, people were not 100% sure (when embedding constructs from same namespace)