Aha, I am now beginning to grasp the problem (having also got round to 
reading about Greek ways of representing numbers: thank you Mr 
Wikipedia). So, for example, in your text you find μβ' and 
you want to record that you think it means one forty-second (rather than 
forty and a half) ... and you feel that 1/42 is a better way of 
representing this than 0.238. (I am still slightly puzzled by this 
preference, mark you -- if it was in fact 40.5 would you want to 
represent that as 81/2?). So either you need a different attribute which 
allows you to express the normalized value not as a decimal but as a 
vulgar fraction (@vulgVal sounds a bit dodgy but might do!), or we need 
to reinstate the cumbersome method we had in P4 of underspecifying the 
datatype of such attribute values (i.e. @value would need to be 
complemented by some global declaration for @valueScheme). I'd regard 
the latter as a retrograde step myself.

Presumably there are different Unicode characters for the greek 
letters-used-as-numbers either existing or proposed?



Gabriel BODARD wrote:
> Lou Burnard a écrit :
>   
>> @value is defined as decimal, as previously noted and should not 
>> therefore be abused in this way. (I find it hard to imagine any real 
>> life application in which 0.333 isn't good enough as an approximation 
>> but that probably just shows I am deficient in imagination)
>>     
>
> Well, I don't especially want to be able to process the fraction 
> mathematically (which is why MathML would be overkill), but what I do 
> need is to be able to display (and index, and search for) <num>ιβ</num> 
> as 1/12 rather than 0.0833·. The decimal is just *not useful* to us in 
> any way, except I suppose for sorting (and it can be trivially generated 
> from the fraction anyway--the reverse is not true). Having to define a 
> set of equivalences in a header for the dozens if not hundreds of 
> possible fractions that can occur strikes me as cumbersome. I suppose 
> something like <num value="12" type="fraction"> could be processed 
> satisfactorily (and is analogous to what is going on in the Greek, where 
> iota-beta actually means twelve and only a diacritic turns it into a 
> fraction), but the @value wouldn't be true, would it?
>
> Of course in DTD-land datatypes aren't enforced, so we could continue 
> with @value="1/12" and not be invalid until we adopt P5 some time next 
> year. :-|
>
> G
>
>