Re: Roman numerals (was Re: How to write Armenian ligatures?)

James Kass wrote:

> Based on the text quoted from the standard, it appears
> that CJK users are welcome to use Roman numerals, but
> Latin script users are discouraged from doing so.

More or less so. But discouraged, not forbidden.

> There are many things I like about the Unicode Standard,
> but assigning a character and then telling people they
> mustn't use it isn't one of them.

Well, it doesn't say "musn't". Rather "shouldn't", and even this 
probably wouldn't express the tone quite adequately. And I'm sure you 
know the reasoning behind this: Unicode was designed so that it can be 
used as a universal code in the sense that character data in any 
encoding can be mapped to Unicode and vice versa, without losing any 
distinction that might be made in some other character code. (This does 
not mean that conversions must preserve such distinctions as letter I 
vs. Roman numeral I if made in the original data; just that conversions 
_can_ be made that way.)

The main practical problem with this is that people who only look at the 
Code Charts, paying attention to headings, glyphs, and names of 
characters, easily get misled into using compatibility characters that 
should normally not be used for new data. They may even think they are 
progressive and logical. After all, if "I" is encoded as ROMAN NUMERAL 
ONE, then grammar checkers, speech synthesizers, indexers, and other 
software can know how to deal with, never taking it as a personal 
pronoun for example. Besides, programs can select a specific glyph, 
different from that of letter "I", for it. There's much sense in such 
reasoning, except that it's not what the Unicode Standard suggests. And 
this could be the start of a beautiful debate, except that it's too 
late. We are supposed to use Latin letters as Roman numerals, and we are 
supposed to use higher protocol levels, such as markup or commands, to 
distinguish between different uses when needed.

I don't think the Unicode Standard can be interpreted as saying that you 
should not ever use the compatibility characters denoting Roman numerals 
in texts written in the Latin script. You might have reasons to do so, 
perhaps just to be able to render them differently in a system that does 
not support higher-level tools. But you should not expect that the 
distinction will be preserved when the data is transferred to another 
system. In practice, another system might not even recognize the Roman 
numeral characters.

The Roman numerals, or at least most of them, were probably not related 
to letters at all but special symbols, originating from counting marks, 
and were later associated with letters. If there were ancient texts that 
make a distinction between, say, letter V and the Roman numeral for 
five, and if such texts needed to be encoded as plain text in digital 
format to a considerable amount, then we would have a case for encoding 
the numerals as independent, non-compatibility characters. In this 
hypothetical situation, the would have to be added new characters, since 
the current Roman numerals have equivalences that would not be 
appropriate for such use.

On the other hand, anything less than such a situation will hardly cause 
any serious reconsideration of the status of Roman numerals in Unicode.

Jukka K. Korpela ("Yucca")
http://www.cs.tut.fi/~jkorpela/