[tei-council] EDW90 proposals (1 of several)

[Syd Bauman]

> 1. Drop the following attributes on teiHeader: ... I have followed
> this advice, and now removed all attributes from the TEI Header
> (other than type). [i.e., creator=, status=, date.created=,
> date.updated= have been removed]

As no one screamed for a resp= on <teiHeader>, I think this is fine.


> I have also added the following text to chapter HD:

As modified (what's in P5 has been updated per Sebastian's
suggestion), this looks fine to me, too. 


> 2. Drop the wscale attribute on <alt> and <altGrp> (presumably
> because the weight attribute should contain a quantity and a scale
> as a single value -- but this is not explicit in Syd's table)

How embarrassing. The note says "see weight= of <alt>", but there is
no entry for weight= of <alt>, as it's the "weights=" attribute of
<alt>. Besides, there's not much really helpful there, so I'll
explain here, then make a quick update to the database.

wScale= serves no purpose in P4 but to differentiate whether the
values in the weights= attribute range from 0 to 1 or 0 to 100. Even
though a value of "75" is unambiguously 75%, a value between 0 and 1
(inclusive) is ambiguous unless wScale= is specified. Thus,
   <alt wScale="real" weights="0 0.5 1"/>
represents odds of 0 (0%), 0.5 (50%), and 1 (100%), whereas
   <alt wScale="perc" weights="0 0.5 1"/>
represents odds of 0 (0%), 0.005 (0.5%), and 0.01 (1%).

However, the suggested datatype for P5 (tei.datatype.probability)
deliberately permits a percent sign in the value itself to perform
this disambiguation. Thus the first example above would be expressed
as either
   <alt weights="0 0.5 1"/>   or      <alt weights="0% 50% 100%"/>


> This seems reasonable in itself. On the other hand <alt> is a
> pretty recondite element which probably needs a lot more thought

Indeed.