Quoting Benct Philip Jonsson <[log in to unmask]>:

> >The big
> >change is that anything that is represented in the IPA by a
> >superscript is represented by ^ plus the sign in question;
> >thus ^h for aspiration, ^j for palatalization and so on.
> >When the superscripted character isn't ASCIIically
> >available, whatever is used for the character in its
> >independent IPA use is used instead; eg ^G for velarization.
> In my unfinished proposals I introduced a
> distinction between _ and ^ for indicating diacritics.
> For example _l indicated 'lateral' while ^l indicated
> 'lateral*iz*ed', _n was 'nasal' ^n was 'nasalized' etc.

Neato. However, implementing something of the sort would be a biggish
upheavel, so I'm only adding it to the possible ideas list for now.

> >I'm sure this scheme perserves a variety of aspects of CXS
> >others find obnoxious.  Tell me, and we can hopefully word
> >out an improvement!
> I absolutely think that the \ and ` diacritics should
> *precede* the character they modify.  In the case of the
> backslash this is the way it is normally used in
> programming languages, and in the case of ` one may
> (at least on a Swedish keyboard, as I'm sure you know)
> inadvertently type e.g. {tā} when {t`a} is intended,
> while {`ta} is fool-proof in this regard.

Probably because of the way letters like 'h' and 'j' are misused as
quasidiacritics in many Latinically written languages, I've always felt that
pseudodiacritics should follow the sign the modify, hard on automatic parsing
software as it may be.

And while frontifying ` would decrease typing pain with retroflex consonants,
it would increase it with rhotic vowels - you'd risk end up writing 'āt'
intending '`at', while the present 'a`t' is fool-proof.

Additionally, both the rhoticity hook and the "claws" of the retroflex IPA
letters go right, so I feel having the ` to the right improves mnemonicity.

> BTW I think that the frequent use of \ should be
> avoided, since it leads to ugly character sequences
> in phonemic transcriptions: /i\/ or /\i/ are equally
> bad in this respect!  I have no immediate suggestion
> for an alternative diacritic, however. Perhaps * ?

I don't much like the backslash for this either, but good alternatives are

Least bad would probably be simply switching * and \ - no-one seems to be much
using the mid-centralizing diacritic anyway. Does anyone else have an opinion
on this?

> >The below mostly reproduces the IPA, but
> >I'm perfectly open to add more non-IPA distinctions. Anyway,
> >this what I have ATM:
> [snip]
> The only thing that really worries me is how the r`/4` distinction is to be
> represented in non-ASCII! :)

Well, solving that didn't figure in my objectives!

> Perhaps the olden click-symbol 'streched c' (U+0297)
> might be locally revived for [4`]!
> BTW there should of course be a retroflex lateral flap
> symbol [l\`] too!

Indeedy! Consider it added.

> >I do not ATM have any better idea than X-SAMPA's _< and _>
> >for implosives and ejectives - I'm open for suggestions.
> I suggest _! or ^! for implosives and _? or ^? for ejectives.

I rather like ^! and ^?. The later is somewhat worryingly similar to ^?\ for
pharyngealized, but we're already upholding some pretty fine distinctions
here, and the question mark is of course not wholly disimilar from the IPA
ejectivizer. They certainly beat the X-SAMPA indications, at any rate, so I'll
add them at least till I find something better.

> >Other consonant symbols:
> >
> >W    voiceless labial-velar fricative.
> >w    voiced labial-velar approximant.
> >H    voiced labial-palatal approximant.
> >H\   voiceless epiglottal fricative (someone tell me what this is!)
> ><\   voiced epiglottal fricative
> >>\   epiglottal plosive (I want to change these two too - suggestions?)
> I used Q\ and P\ as ugly hack alternatives in my CXStoHTML perl script,
> since it is my absolute opinion that < and > should be avoided in something
> that is to be put inside an HTML document.

I'm still not clear what epiglottals _are_. However, I suspect that have very
little to do with any sound reasonably denoted with a 'p', capitalized and
diacritic'd or not. I think I rather prefer Q\ and q\, since they at least
vaguely suggest something far back, and I assume the "-glottal"
of "epiglottal" also does. I'll be using that for now.

I'd very much appreciate if someone took the time to explain epiglottals.

> >s\   voiceless alveopalatal fricative
> >z\   voiced alveopalatal fricative
> >s\!  palatoalveolar click
> >l\   alveolar lateral flap
> >x\   simultaneous S and x
> >5    velarized alveolar lateral approximant
> >
> >Affricates and double articulations may optionally be
> >inclosed in { } to disambiguate.
> An old suggestion of mine.  I used the same in my proposal

Yep. Should've mentioned you in the credits, I guess!

> >Alternatively, affricate or
> >double articulation may be assumed, and clusters separated
> >by '-'. Note that normal parentesis and square brackets
> >retain their IPA functions!
> >
> >Vowels:
> >
> >|  i  y        i\ u\    M  u
> >|      I  Y    I\ U\     U
> >|    e  2      @\ 8     7  o
> >|                @
> >|      E  9    3  3\    V  O
> >|       &        6
> >|        a  &\          A  Q
> What about [8\] for [2^w]?

Consider it added, as an "other vowel symbol" in analogy with the "other
consonant symbols.

> >Superagementals:
> >
> >'    Primary stress
> >,    Secondary stress
> >:    Long
> >;    Half-long
> >;\   Extra short
> >.    Syllable break
> >|    Minor (foot) group
> >||   Major (intonation) group
> >
> >For tone, I don't have any improvements on CXS to suggest
> >ATM. However, if '<' and '>' are freed up, I'm thinking they
> >could be used to enclose tonal info. Eg, [ma<TMH>] would be
> >the syllable "ma" with an obnoxious extra high-mid-high
> >contour tone on. Since ! and ^ have been hijacked, it would
> >also allow us to use <!> and <^> for downstep and upstep.
> Good Idea.

Since '<' and '>' are now free, the above is official.

> May I also suggest the not unprecedented use of numbers 1-5 for
> tone annotation on the principle:
> <5>     extra high
> <4>     high
> <3>     mid
> <2>     low
> <1>     extra low
> If you will it is a way of transcribing Chao-style tone letters.

Chao-style tone letters? I'm unfamiliar with that.

> It also allows for annotation of contour tones like this:
> <15>    rising
> <51>    falling
> <45>    high rising
> <12>    low rising
> <454>   rising-falling

I can't see any reason we couldn't afford some reduncany here. So let's allow
either BLMHT or 12345.

> Additionally one may use # to indicate numbered tones, like
> /ma#1 ma#2 ma#3 ma#4/ for Mandarin, or /anden#1  anden#2/ for Swedish.

Fine for phonemic transcription, I guess, and as long as we've got at most ten

> >Diacritics:
> >
> >_0   voiceless (zero)
> >_v   voiced
> >^h   aspirated
> >_o   more rounded (using lowercase lessens risk for confusion with _0)
> >_c   less rounded
> >_+   advanced
> >_-   retracted
> >¨    centralized (did not seem to be any reason not get rid of the
> underline)
> NB ¨ is not a valid ASCII character.  I take it you mean ".


> >*    mid-centralized
> >=    syllabic
> >=    non-syllabic (can't think of any symbol that need distinct syllabic
> and
> >non-syllabic diacritics!)
> >`    rhoticity
> >_¨   breathy voiced
> Same comment as per centralized above.
> May I suggest _h as an alternative?

It's a typo for _", inspired by the IPA diacritic. I like it better than _h,
but I guess a case could be made it ought be _h\ or ^h\. But I'm keeping _"
for the moment. _h being unclaimed could of course be an optional alternative.

> [snip]
> >_`   no audible release (similarish to IPA diacritic, and I hate brackets
> used
> >for non-brackety purposes)
> Agreeissimo!
> >The ^X = superscript X convention may also be used for
> >writing explicitly falling or raising diphthongs: [a^i],
> >[^uo].
> NB that {u^o} might be ambiguous with your 'lowered'
> diacritic.  E.g. German _au_ is phonetically {ao}.

I _did_ write [^uo], and ^u isn't defined as a diacritic! You're right in
substance, however; the scheme is presently ambuiguous here.

Hm, by my own rules, the nasal and lateral release should be ^n and ^l, which
frees up _l for lowered, which is mnemonically to be prefered, methinks. Then
we'll also have _r for raised. Which nicely solves the ambiguity with ^o, too.

I'd say that German diphthong is closer to [{Ao}], tho.