On Thu, 10 Jun 2010 13:16:58 -0700, David Peterson <[log in to unmask]> wrote:

>[...] I
>like my romanization systems to be mostly phonetic so that users
>don't have to memorize phonological rules just to pronounce the
>language (if I were going to have people go that trouble, why not
>just use the orthography everywhere--especially with the new
>font-embedding technology that, for example, Arthaey Angosii has
>used to great effect?).

Eh, different strokes.  I like 'em to be phonemic (in cases where
phonemehood is uncontroversial) 'cause allophonic detail can be distracting
and I mostly just think the underlying forms are cleaner.  If pronunciation
is that paramount, just use IPA.

>Speaking of which, I had an idea that I haven't fully fleshed out
>yet (too many variables) about a single, unified romanization
>system that we could use for everything. The reason is that unless
>one's language is spoken by fictional speakers that have access
>to the Roman alphabet, a romanization is not an orthography.
>As such, it's nothing more than a tool, and it'd be nice if there
>was a single way to apply it to every language.

I've thought about this a little in connection to my language generation
program (for which I'll eventually want a component much like this), and I
mostly agree.  But there are a few broad choices I wouldn't want to make for
all users of the Unified Romanisation System, and I'd want to leave them as
parameters.  One of these is depth (do you write phonemes or allophones?),
per above.  

The other big one is digraph-friendliness and diacritic (or nonbasic
letter)-friendliness.  These go together for the obvious reasons: if you've
more than 26 plus a few sounds you'll need one or the other.  That said,
there're technological reasons you might want to eschew diacritics; there're
more or less phonotactic reasons you might want to avoid digraphs (the more
freely [sh] can occur, the worse <sh> for /S/ is); and there're aesthetic
reasons for either.  I'm not sure yet whether I also want to endorse being
able to say no to both, in the cases where that's feasible...

Actually, you know what, in that light it would be more agnostic to make the
standards go the other way.  Don't say "if you have [S] you should spell it
such-and-such"; say instead "if <sh> is a digraphic unit of your
romanisation it should represent [S]" (well, some coronal fricative anterior
to [s], but that's not the point here), and also "if s-hachek is a unit of
your romanisation it should represent [S]".