Detail #53: Grammatical and pragmatical length

I suspect this violates some universals pretty strongly, which pleases me quite a bit.

Imagine a system where length is phonemic, but does not distinguish lexemes. Instead, length comes in patterns such as

sLss, LsLs, LLss, LLLs, ssss, LLsL

where L = long, s = short. These are applied to individual phonemes. Different patterns signify different grammatical and pragmatical information. Example patterns - just for illustrative purposes - given with Kleene-algebra notation. Some of them may span several words, some of them are applied sentence-initially, some are applied from the start of a noun.

verbal and sentence-initial patterns:
LLssL(s*) = yes/no-question
(sL)* = indicative, unmarked
s* = indicative, unmarked
s(LL)s* = imperative
s*LL = if interrogative particle is included, marks that the listener is repeating what he thought he heard in order to get confirmation he heard right

noun-markers:
s* = indefinite nom or acc
sLsLs* = definite nom < barely distinguished
ssLLs* = definite acc < barely distinguished
Ls* = fronted direct object

adjective markers:
LLLs* = intensive
s* = default
s*LL = verb-like, but restricted in not being able to take verbal patterns

Certain weird things can happen for phonemes that do not permit lengthening, such as the lengthening shifting one step to the right, or things along those lines. Finally, some particular patterns may trigger the appearance of extra phonemes just to carry length, usually /e/ or /n/, except after /k/, which triggers /u/ instead.