Re: Expert opinion on <context-group> uniqueness

["Matthew Lovatt" <matthew.lovatt@oracle.com>]

Sorry Yves, I have not used context-group so far.
Are you thinking of someone else in the group?

I have been looking at the spec and I think there is an ambiguity in the
definition.

According to the spec, each context group must be uniquely named in the
file.

It then describes how a default context group can be defined at a high level
in the hierarchy which is implicitly inherited at all child levels, but does
not explain how the overrides behave

For individual sub elements , default group can be overridden at lower
levels.

But for this to work, the override group MUST have the same name as the
default group, or how will the parser realise that the groups are identical?

Therefore,
1) Multiple context groups with the same name MAY exist but only in
different levels of the hierarchy
2) Defaults and overrides MUST have the same name - matching context names
are used to define parent child relationships

At least that's what I think.

I have not CCed Asgeir, as others in the group may not agree with this
definition.

If consensus in the group, please forward to Asgeir.

If there is a real ambiguity, do we simply need an explicit example of using
default /override, or do we need to modify the spec again?


Mat
----- Original Message ----- 
From: "Yves Savourel" <ysavourel@translate.com>
To: "'Asgeir Frimannsson'" <asgeirf@redhat.com>
Cc: "'Matthew Lovatt'" <matthew.lovatt@oracle.com>
Sent: Tuesday, May 03, 2005 3:47 PM
Subject: RE: Expert opinion on <context-group> uniqueness


> Hi Asgeir,
>
> To be frank, I haven't used the <context-group> structure much (and not
lately).
> Looking at the definition it seems you can't have the same name within the
same <file>.
>
> But you say:
>
> > If B is correct (which I think it is, by looking at the examples you and
> > others have used in XLIFF TC presentations etc), how then about PO and
references?
>
> So, I must have wrote something that indicates the answer is B. (and that
may be incorrect).
>
> The best person to answer you question would be probably Matt (I CCed
him). I think he is using extensively <context-group>.
>
> Any thoughts Matt?
>
> Kenavo,
> -yves
>
>
>
> -----Original Message-----
> From: Asgeir Frimannsson [mailto:asgeirf@redhat.com]
> Sent: Tuesday, May 03, 2005 1:10 AM
> To: Yves Savourel
> Subject: Expert opinion on <context-group> uniqueness
>
> Hi Yves,
>
> Another issue we've been discussing up and down is the 'name' attribute of
<context-group>. In the XLIFF 1.1 Specification, under
> the description of the <context-group> element, the specification says:
>
> "The required name attribute uniquely identifies the <context-group>
within the file"
>
> We've been discussing the meaning of this, and have two distinct opinions:
>
> A) Each <context-group> element must have a unique name. You can't have
two <context-group> XML elements within the same <file> with
> the same 'name'
> attribute.
>
> B) Each collection of <context-group> elements must have a unique name.
> Meaning you can re-use the same 'name' attribute for a <context-group> at
a different hierarchical level in your XLIFF file.
>
> If B is correct (which I think it is, by looking at the examples you and
others have used in XLIFF TC presentations etc), how then
> about PO and references? It is common that each translation unit in PO
have multiple references, e.g.:
>
> #: myfile.c:23, myfile.c:45, myotherfile.c:12 msgid "hello world"
> msgstr ""
>
> our proposed approach is to use context groups, similar to:
>
> <context-group name="po-reference#1">
>   <context type="sourcefile">myfile.c</context>
>   <context type="linenumber">23</context> </context-group> <context-group
name="po-reference#2">
>   <context type="sourcefile">myfile.c</context>
>   <context type="linenumber">45</context> </context-group> <context-group
name="po-reference#3">
>   <context type="sourcefile">myotherfile.c</context>
>   <context type="linenumber">12</context> </context-group>
>
> Here we add #n to the context-group name to ensure a unique name at the
same hierarchal level, but I feel this is a bit of a 'hack'.
> They should be member of the same named context group, i.e 'po-reference'.
Any thoughts?
>
> (Note: I sent a similar mail to the xliff-comment list 3 weeks ago, but no
> reply:
> http://lists.oasis-open.org/archives/xliff-comment/200504/msg00000.html
> )
>
> [feel free to cc the xliff-tools mailing-list if appropriate]
>
> cheers,
> asgeir
>
>