Peter, > > Why you would feel that such user sense of the characters they > > are using is belied by your analysis of the shape of the hooks > > used in the IJAL font is beyond me. > > I'm sorry I wasn't clearer. I was not referring to their status in terms of > defining characters. I was *only* referring to the typography -- their > shape. Their status for purposes of defining characters was what I was > eventually trying to get to, but that statement was not meant to reflect a > conclusion.
O.k., I understand the differences in typographic tradition and acceptable variability. I was just concerned this was prematurely (and rapidly) headed towards the "what additional characters are needed to encode these?" question, when, in fact, all the characters needed are already encoded. > On revisiting Pullum and Ladusaw's comments on "Polish hook", however, it > all became clear: "Boas et al. (1916, 8) recommend the use of a centered > subscript rightward hook under vowel or consonant symbols as a nasalization > diacritic." The criteria provided leave a fair amount of room for > variation, and clearly the variations I was seeing all fit. > > So, I agree with your statement, > > > Very simply. They use vowels with ogoneks to represent vowels with > > ogoneks. > > but I would point out to the typographers out there that not all ogoneks > have to look exactly the way that Adam described when the usage context is > phonetic transcription (though, for Polish, I'm sure they would). Exactly. Although even in the case of Polish, what John and Adam are saying seems to be a prescriptive norm for Eastern European typography, which is, of course, often observed in the breach. Those same American journals using funky hooks for ogoneks for Onondaga or Lakota or Dogrib might just as well have Polish data in them using the same fonts for those vowels with ogoneks. > > John's comments and Adam's paper (both absolutely correct in relation to > typography of European languages) misled me. > > The are two widespread conventions in Americanist orthographies > > for nasalization: ogonek (or "Polish hook") under vowels (or semivowels) > and > > tilde over vowels (or semivowels). > > There is another convention, admittedly far less widespread: cedilla. > Brewster and Brewster used this, and I have seen vowels with cedilla used > by various linguists in recent publications (e.g. Givón 2001), apparently > for nasalisation (in some samples I have come across, the notation is not > explained). I don't think this is an accepted convention. See Pullum and Ladusaw again. I think those instances where you find a linguist publishing using vowels with cedilla for *nasalization* are results of either: a. confusion by the user regarding the direction of the hooks, or b. symbol substitution when using a font that had the cedilla forms but not the ogoneks. The cedilla with a vowel is an Indo-Europeanist convention for a more open vowel (= IPA U+031E) -- cf. Brugmann, 1904. It may well be that some materials using vowels with ogoneks for nasalization ended up typeset with fonts which had the vowels with cedillas, and the compositor/editor decided to just go with the substitutions. Such substitutions in fonts when faced with difficult phonetic orthographies are not uncommon in 20th century linguistic publications, although they are thankfully becoming less of an issue now as fonts and font technology improve. > > No, they map vowel+ogonek (whether as a precomposed form or as > > a vowel + combining ogonek sequence, which would be canonically > > equivalent > > Ken, given that existing precomposed vowel+ogonek forms have canonical > decompositions, whereas characters with palatal hook and retroflex hook do > not, I'm wondering whether your recommendation regarding atomic encoding of > such character modifications extends to characters with ogonek. No. Given the precedent we have, where letters with cedillas and letters with ogoneks are given canonical decompositions, those two (mostly) attached diacritics fall on the nether side of the thinly drawn line between entities which are better handled as unanalyzed units and those which can simply be represented by existing letters in combination with existing combining marks. > There are a > total of five vowels (aeiou) for which Unicode has precomposed characters > with ogonek. There are others that are used with ogonek in phonetic > transcription (potentially, any vowel symbol can take ogonek, though I've > encountered a much more limited set after looking at quite a number of > sources). Should these be encoded using U+0328, or should atomic characters > be proposed? Using U+0328. Note that the example you posted also had an h-ogonek, so the usage is not limited to vowels, per se. (Although that particular entity itself is a little bizarre, since you cannot really nasalize a voiceless glottal fricative. I suspect that a closer phonetic representation of those symbols should have been <U+0266, U+0328>, using the h-with-hook for the breathy-voiced glottal fricative, much more likely to occur in an intervocalic nasalized environment.) --Ken