Stefan Majewski wrote:
> James Cummings wrote:
>   
>>> The P4 encoding I am trying to convert to P5 goes like this:
>>>
>>> <gap desc="initial not executed" reason="unfinished"/>
>>>
>>> Should this now be <gap reason="unfinished"><desc>initial not
>>> executed</desc></gap>?
>>>       
>> Yes, definitely!  The reason for this is the war on free text-bearing 
>> attributes.  You'll see that most attributes now in P5 have a datatype 
>> and don't take just free random CDATA.
>>     
>
> Which in my opinion has disadvantages as well. We use our texts with a 
> variety of tools, not all of them are very good in omiting particular 
> elements when doing a full-text search. This is not an issue with eXist, 
> it is simple there. But for example with Xaira, you may not only exclude 
> some elements -- which by the way works quite well -- but also use 
> regions which makes Xaira ignore the exclusions.
I am not sure I understand what you are saying here -- with Xaira you 
can exclude elements from indexing (so their content won't appear in a 
word search), and you can also use regions to define areas to be 
searched in irrespective of their indexing policy. The two facilities 
are complementary. However,  this is probably a discussion for the xaira 
list, not here.
>  While I see that the 
> reasons for having elements rather than attributes, may be stronger we 
> still wanted to keep the description as an attribute as we preferred not 
> to have information on the text that is not really text itself to be 
> encoded as text() node below //body. 
The Council discussed this concern several times and it has been a theme 
of our "what's new in P5" talks for the last few years -- there are 
several places in TEI now where you can't just remove all the tags and 
recover "the unmediated text". <choice> is one, and the many places 
where text attributes have become child elements are another. Happy to 
have the argument again, but I doubt if it would persuade anyone not 
already persuaded...


> We just put a @desc  in our  own 
> namespace, which I hope should be fine, or at least not too abusive.
>
>   
Absolutely fine -- this is the way to do it.


> So we are doing something like:
>
> <gap voice:desc="initial not executed" reason="unfinished"/>
>
> best,
>
> Stefan
>