Re: [office] describing and checking the relation between identifiers and references

[Jos van den Oever <jos@vandenoever.info>]

On Monday 19 October 2015 18:52:32 Jos van den Oever wrote:
> Hello all,
> 
> I have added a proposed addition to the specification for issue 3887.
>   https://issues.oasis-open.org/browse/OFFICE-3887
> 
> (sorry for all the svn commits, i'm learning to cope and I'm spoiled by git)
> 
> 
> https://tools.oasis-open.org/version-control/svn/office/branches/v1.3_OFFICE
> -3887/v1.3/identifiers-and-references.xml
> 
> lists many of the relations between identifiers and references in ODF 1.3
> (and 1.2)
> It is written in a simple XML format that is hopefully readably by you all.
> If you require clarifications, please tell me.
> The format should be simple enough so it can be used for code generation by
> implementors.
> 
> https://tools.oasis-open.org/version-control/svn/office/branches/v1.3_OFFICE
> -3887/v1.3/iar2sch.xsl is an implementation for checking the ids and refs
> with Schematron and Saxon. It contains instructions on how to use it.
> 
> Together with the Relax NG, this set of relations should help to make it
> easier to create new implementations and improve existing ones.

I think I found an issues from writing the 

19.874.20<text:index-title-template>
"The text:style-name attribute specifies character family styles."

However, the attribute text:style-name on <text:index-title> says:

19.874.19<text:index-title>
"The text:style-name attribute specifies section family styles."

Contrary to the specification, LibreOffice uses a paragraph style in text:index-
title-template/@text:style-name. It does use a section style for text:index-
title/@text:style-name. It makes sense that the style family for text:index-
title-template/@text:style-name is at least paragraph.

LO uses the style from on a paragraph in <text:index-title>. This seems 
logical.

So you see that careful reading of this new file and using it on existing files 
can lead to interesting finds. Please have a look through the file if you want 
and/or test it. There are still quite a number of id/ref pairs that can be 
added.

We should discuss if we want to make this list of ids/refs informative or 
normative. Like with the RNG schema, the contents of the file could be use to 
create parts of the specification.

Cheers,
Jos