Tightens onset covering grammar, converting the onset rule to four CDRewrites:
word-initial cluters
mandatory splitting of tl and dl
optional, non-zero cost splitting of other MCLs
mandatory onset parsing for remaining MCLs, labiovelars, and prevocalic singleton consonants.
Thus MCL splitting is disfavored except when required by the meter. The costs in variable.grm are increased to reflect this.
As a side effect this enforces the point in #55, part 1.
A note on terminology: I hypothesize, on the basis of the behavior of bijugo, that [j]-gemination may be blocked when the [j] comes to be intervocalic by compounding or prefixation (note also that the [j] in prae + vowel-initial stem is usually not geminate either in Vergil, according to Cser). Linguists refer to this syndrome---a rule that doesn't apply in environments derived by other processes, usually morphological ones---as derived environment blocking. (Most rules apply in both derived and underived environments; a few appear to apply only in derived environments.) There's not much we can do about this without actually performing a morphological analysis along with the phonological analysis we're already doing.
As an unrelated change, also reflows lines at 80 chars.
Tightens onset covering grammar, converting the onset rule to four CDRewrites:
Thus MCL splitting is disfavored except when required by the meter. The costs in variable.grm are increased to reflect this.
As a side effect this enforces the point in #55, part 1.
A note on terminology: I hypothesize, on the basis of the behavior of bijugo, that [j]-gemination may be blocked when the [j] comes to be intervocalic by compounding or prefixation (note also that the [j] in prae + vowel-initial stem is usually not geminate either in Vergil, according to Cser). Linguists refer to this syndrome---a rule that doesn't apply in environments derived by other processes, usually morphological ones---as derived environment blocking. (Most rules apply in both derived and underived environments; a few appear to apply only in derived environments.) There's not much we can do about this without actually performing a morphological analysis along with the phonological analysis we're already doing.
As an unrelated change, also reflows lines at 80 chars.