I have just stumbled across a potentially serious problem with the
definition and use of nonspacing character mnemonics in RFC-CHAR. Previously
I thought this wasn't a problem, but I was wrong.
This is best illustrated by an example. The mnemonic "? is a combining
tilde accent. When this character appears it is understood to be positioned
above the character it is adjacent to.
The question then is whether it applies to the character it preceeds or the
character it follows. According to Unicode Version 1.0 volume 1, "all such
(non-spacing) follow the base character which they modify" (this is on
page 17 in the second paragraph). However, these non-spacing characters
appear in T.61-8bit, and according to T.61 accented letters are "represented
by a sequence of two bit combinations. The first ... representing a
diacritical mark. The second ... representing a basic Latin character." (see
section 4.1.3.b in T.61)
These two usages are clearly at odds with each other. The only solution I
can see is to define two sets of non-spacing mnemonics so that the cases
where they preceed and the cases where they follow the character they modify
can be distinguished. Note that there probably won't be equivalents for
these additional mnemonics in Unicode and possibly not in 10646.
Non-spacing characters appear in ISO 5426, ISO 5427, ISO 6937, T.61, and
T.101, to name a few.
One additional minor glitch -- the definition of '", the mnemonic for
double acute accent, appears to be missing from the current draft.