[tei-council] policy tested (was "Re: [TEI-L] <measureGrp> and <text>")

[Lou Burnard]

This posting demonstrates rather well the slipperiness of the concept of 
"corrigible error". We can all agree that a content model of 
model.measureLike or <text> is plainly wrong for this element. I can 
certainly fess up to the fact that I wrote  "<text>" when I should have 
written "text".  So on the face of it the obvious correction is to 
replace <rng:ref name="text"> with <rng:text> since that's what I 
intended: just as if I had written "txet" instead of "text".

However, there are at least three other and arguably better ways of 
correcting this error:
(a) one should be consistent with other uses of <rng:text> and  change 
the content model to reference macro.xtext
(b) one should be consistent with other uses of the element name suffix 
"Grp" and disallow "text" completely
(c) one should rename the element to something else and allow text (or 
macro.xtext)

Given such shifting sands (and I haven't even started on the question of 
whether the attribute classes are right) I think the "treat it as a 
typo" solution is probably the safest course, even though it does break 
the letter of the Birnbaum law about not making changes in the schema.


Syd Bauman wrote:
> Patrik Nyman just posted about what I would call a corrigible error
> in P5. The content model of <measureGrp> refers to the <text>
> element, where it is quite clear that what Lou intended was the
> 'text' RELAX NG pattern, and I think if it is there at all[1] should
> be the "macro.xtext" pattern.
> 
> The problem here is that our announced policy was not to make any
> changes to P5
> * that affect the schemas for the 2008-01 release, or
> * that break backward compatibility ever.
> 
> Personally, I think both of these are too restrictive, and that we
> should have an explicit and publicly stated policy of fixing
> corrigible errors ASAP regardless of whether or not they break
> backward compatibility. But that is not the situation now.
> 
> 
> Note
> ----
> [1] Lest anyone think that my belief that 
>         if character content is allowed inside <measureGrp>, then <g>
>         should be allowed inside <measureGrp>
>     implies that I think character content should be allowed inside
>     <measureGrp>, I will say again what I said repeatedly over the
>     summer: <measureGrp> should *not* allow character content. If we
>     had wanted the "box to hold compound measurements" to contain
>     text, then it should not have been named <measureGrp>, but rather
>     <measureSet> or <measurements> or some such.
> 
> _______________________________________________
> tei-council mailing list
> tei-council at lists.village.Virginia.EDU
> http://lists.village.Virginia.EDU/mailman/listinfo/tei-council