A brief reply to a fragment of Tobias' larger work:
>* Some fractions are defined in the ISO lists, like 1/7, but others are
> not, like 5/11. Also, there is an entity representing `c/o'
> (&incare), but none for the similar `a/c'. Would you suggest:
> (1) using ISO &frac17 for the one and Non-ISO &frac5.11 (or so) for
> the other, resp. &incare for the one and (say) &aoverc for the
> (2) in order to achieve more uniformity (in rendering, e.g.), use
> Non-ISO entities for all of them -- the replacement text for those
> in ISO could be their ISO entity name again, but would not have to
> (3) for the fractions, use a completely different scheme, like <num
> value="5/11" type=fraction> or even (please no) <num
> value=0.45454545 type=fraction>?
I have some experience here from the British National Corpus.
1. In the BNC we supplemented the ISO entities with additional ones so
that we covered all fractions having a denominator of nine or less
and which was not a whole multiple of the numerator. (Not all were
used; of the possible ninths, only &frac19; appears in the corpus,
for example.) Anything not fitting this pattern because of denominator
> 9 or non-numeric numerator or denominator was, for reasons of
expediency, left as one_thing/another. In particular, we didn't do
anything special with c/o and so on.
The trouble with trying to construct a closed set of entities is
to represent vulgar fractions that you have to know a priori
what is going to appear in your document instances. Usually, you
don't. The size of set of vulgar fractions is a large infinity;
attempting to enumerate it with entities is a mug's game. If you
use entities for any fractions, you need an escape route. The BNC's
was simple but not very useful; one could do better.
2. I'd advocate using the ISO entities where they exist. That's why the
BNC stopped at denominator nine: that way, our entities had the same
appearence as ISO's.
3. I like the element suggestion best, perhaps with the following
<number type=fraction num=5 den=11 value=0.45454545>5/11</number>
What entities you have can expand into these elements, which have content
because my experience is that people just hate seeing their content
disappear into attribute values. (Duplication is OK; theft is a no-no.)