There are some possible downsides though to the minimalist, "
ca'e /S/ and /yS/ are now allophones" approach:
- The tests run were based on specific words; there is a slight possibility that there is actually a form somewhere that could break up, and unless we come up with a mathematical proof that this change can't make any crazy pathological word break up, word boundary ambiguity might crop up somewhere. It's better for the JVS algorithm and parsers to be equipped to catch this error.
- Syllabic consonants just look ugly, pe'i. They make syllabic resolution difficult; and I can somewhat imagine a potential learner going "so the only way to "safely" make a loan word is to crunch all these consonants together*? ... I'm out". Then again, I entered {briii} to JVS, so I don't know how seriously anyone would take my opinion :p
* or, if la .guskant.'s month/day names catch on, "do all the month/weekday names look this way?" **
** [incidentally if it weren't for the syllabic consonants I would favor them over {pavma'i} et al, which rely on numeric rafsi forms with additional consonants to learn]