From: Peter Constable (petercon@microsoft.com)
Date: Wed Feb 18 2004 - 19:44:35 EST
> From: unicode-bounce@unicode.org [mailto:unicode-bounce@unicode.org] On Behalf
> Of Michael Everson
> >- If such usage should arise, how would our
> >decision to encode [/] affect how we decide
> >about [¯] or [Å®]?
>
> Ignores that parentheses and hyphens are already encoded as subscripts.
Well, *don't* ignore the fact that I was not talking about parentheses and hyphens; I was talking about tilde and right arrow (U+2192).
> >The point is, we don't (or, at least, shouldn't)
> >just encode things because we saw them being
> >used. We should establish principles (however
> >formally or informally stated) that we use to
> >guide our decisions. *That* is why Ken cares
> >about a possible subscript tilde.
>
> The boundary is what people need to encode, not
> whether it fits in with one particular linguist's
> view of what merits veto.
Sorry, that doesn't cut it, IMO. You know all too well that "what people need to encode" is open to lots of interpretations. When someone comes along and says "we need to encode Devanagari ksha" or "we need to encode ch", you, Michael Everson, will respond saying, "No, we do not need to encode those things," and you make that decision by applying some principle. Unless you come up with some principle other than "these people need it", then you'll be inviting those wanting ksha or ch or other such things to ask for the same, and with equal justification.
We encode super/subscripts for speech sounds in phonetic notation, I think, because
- the usage is conventional across a significant and identifiable community,
- because the super/subscripting is semantically significant within those notational conventions, and
- because it is a convention of finite scope: it involves particular symbols all used in the same way -- to indicate some modification of another sound or some type of secondary aspect of articulation -- and is not part of a richer set of notational devices that would require higher-level layout constructs in order to represent adequately.
In the case of x and particularly / it is not entirely clear that the last point is true. Can we say that it is a natural extension of the notational device that does not encroach upon some higher level of representation? This especially is in question. It's one thing to say
"this laryngeal has a particular quality of being coloured by 2nd and 3rd formants that correspond to the vowel a"
That's all part of simply describing a speech sound. When you start using notation that means
"this laryngeal has a particular quality of being coloured by 2nd and 3rd formants that correspond to *some vowel that might have been e or might have been o*" (or "...that was e in some contexts and o in other contexts").
then you've gone beyond merely describing a speech sound. You're adding meta information giving commentary about the limits of knowledge and the range of possibilities for what is unknown, or identifying that there was some range of variation that resulted in one of a set of possible values being used, or something of that nature -- whatever is meant, it goes *some* distance beyond simply representing a speech sound.
And that leads to questions: Is this *too* far? Do we expect a clear boundary beyond this, or are we going to find the limit being gradually pushed farther and farther until we find ourselves allowing in all manner of cruft?
Ken isn't worried merely about a possible tilde; he's worried about what the *principle* is that encoding this would would imply by precedent. And I think it's a valid question to be asking.
I'm not saying you cannot come up with a way forward for the x and / to be encoded, but it seems to me you've got to make clear what the operational principles are that guide this decision and that can guide decisions of a similar nature in the future.
Because whether it's the tilde or something else, it will more than likely come up again.
Peter
Peter Constable
Globalization Infrastructure and Font Technologies
Microsoft Windows Division
This archive was generated by hypermail 2.1.5 : Wed Feb 18 2004 - 20:22:44 EST