[sword-devel] osis2mod and container elements

[Chris Little]

DM Smith wrote:
> Chris Little wrote:  
>> It looks like I've finally found _one_ thing OSIS can't encode, related 
>> to this. OSIS is unable to encode a paragraph that spans consecutive 
>> books (because <p/> can't contain <div/>).
>>
>> There is actually a case in which this is needed, across the 
>> 1Kings/2Kings boundary in the Hebrew.
>>   
> One can use the milestoned version of div. Mechanically, it is valid xml.

"<p><div/></p>" isn't valid in OSIS 2.1.1.

> IMHO:
> This would be bad OSIS.
> 
> The problem with milestones is that it allows for free-for-all documents.
> The only reason to have milestones is to have two different markup 
> systems at the same time.

In the case I mention here, there is paragraph identification, but those 
paragraphs in one instance cross a book boundary. It's another case of a 
piece of real, inherent textual structure competing with a somewhat 
arbitrary piece of structure that happens to over lap in a way not 
permitted by strict containment.

> My rule of thumb:
> If the element cannot contain another, then using milestones to work 
> around it is wrong.
> 
> I think we need to have a smart OSIS validator that checks two things 
> that an XML validator cannot.
> 1) Ignoring BSP, does BCV follow strict containment rules.
> 2) Ignoring BCV, does BSP follow strict containment rules.
> 
> Basically, if one were to get rid of Book, Chapter and Verse divisions, 
> all milestoneable elements should be able to be converted to their 
> container form and the document be valid OSIS.
> Likewise, if one were to get rid of BSP elements and convert milestoned 
> versions of Book, Chapter and Verse to their container form, the 
> document should be valid OSIS.

That's fine, but in this case, with a Hebrew Bible, certain book 
divisions are arbitrary, being based not on the textual structure so 
much as on limits of the physical medium itself (i.e. scroll length).

I ended up simply making every book begin and end with a paragraph 
boundary, but at the same time, I'm aware that I'm not encoding what is 
actually in the text.

--Chris