Record of signifcant changes in zasni gerna
All sa-handling removed. I find neither the selmaho nor the construct version of sa-handling very satisfactory, and it adds horrible complication to the grammar, so I'm removing it for now.
The construct free* allowed wherever indicators are allowed. It doesn't make sense to restrict it on the grounds that it can be too long and thus interrupt certain tight constructs, because indicators themselves can be made as long as desired. The only required exception is that, for obvious reasons, "(number / lerfu-string) MAI#" can't be attached to PA# and BY#, and vocatives can't be attached to COI# and DOI#.
Extended its grammar to that of indicators, i.e. it is allowed after any word.
Simplified grammar along the lines of internal grammar of tags.
Separate number and lerfu-string to consist of digits-only and lerfu-only repectively. The occasions when we want them separate are much more common than when we need them together (e.g. "abu za'ure'u cusku" or "ci by .e re cy", etc.) When we need to use digits in a lerfu-string, we can use pabu, rebu, etc. We don't really need lerfu as part of a number. Dimensioned numbers are normally sumti, so "lo ci my" is perfectly usable, and makes more sense, instead of "li ci my".
Merge CMENE with BRIVLA. Small cost and large benefits.
Merge with UI (therefore GOhA merges with BRIVLA and BIhI merges with JOI).
Merge with JOI
Removed capitals as marks for stress, because they make the grammar unnecessarily complicated. Although they don't add any comparable complication, commas and digits also removed, as they should not be part of the basic grammar (eventually they belong in a particular implementation as conveniences).
Eliminated from fuhivla, except for the hyphen of type-3. A 'y' is also allowed in front of the hyphen, so that they are not even required there. This has the additional advantage that 'l' is never needed as hyphen.
Restricted to full erasure. Exceptions for NIhO/TUhE/LU/TO can cause weird behaviour.
Is .y a word?
- .y is a word, and it is used to indicate hesitation between most word pairs. Like all words, however, it must obey some syntax rules, so it cannot be used for example to indicate hesitation before SI, BU and ZEI or after ZEI, ZO and ZOI.
- .y is a word, and it is used to indicate hesitation between words anywhere except before BU. Unlike all other words, it cannot be erased with SI, form lujvo with ZEI, be quoted with ZO or used as delimiter with ZOI.
- .y is not a word, it can be used as hesitation anywhere between words. cmavo can be of form V, VV, VhV, CV, CVV, CVhV, CY, YhY and YCV, but not of form Y. The special cmavo .ybu is the only one of form YCV. Forms YhY and YCV have only one cmavo each, both in selmah'o BY.
- .y is not a word, it can be used as hesitation anywhere between words. The name of the letter "y" is "tersei bu", similar to the name "denpa bu" for "."
I am going in this grammar with option 4 as the one that seems simplest and most reasonable. Option 3 was the one previously implemented, but I find it creates ugly complications in the grammar.
Eliminated except for the simplest case with gek. Elimination of selma'o CEhE, PEhE, NUhI and NUhU.
Eliminated for sumti and bridi-tail connection grouping. They don't work for all cases of grouping, and since they are forethought they are redundant to forethought connectives anyway.
Merge operator and JOhI with joik-jek. Eliminate selma'o BIhE. See MEX grammar proposal.
Use ME instead of NUhA to convert operator (including connectives) to selbri. Selma'o NUhA eliminated. There is no reason to have to remember yet another converter when ME can do the job. (Eventually "nu'a" can be included in selma'o ME to preserve backward compatibility.)
Allow both connectands to be of the same kind, thus:
- gek sentence gik sentence
- gek bridi-tail gik bridi-tail
- gek terms gik terms
- gek sumti gik sumti (not: gek sumti gik sumti-4)
Allow tuhe-tuhu sentences and sentences with elided selbri with NOI and NU
Move up to selbri level.
Make it left-grouping, like everything else.
Eliminate. Its function is too specific. It doesn't work for example to say "or viceversa", or to viceverse tagged sumti.
- JOI, JA, A, BIhI, VUhU, JOhI -> JOI
- FEhE, MOhI, NAhE -> NAhE
- KI, BAI, PU, ZI, ZEhA, FAhA, VA, VEhA, VIhA, TAhE, ZAhO, CAhA, CUhE -> BAI
- LA, LE -> LE
- FUhE, UI, CAI, NAI, DAhO, FUhO, RAhO, GAhO -> UI
- ME, NUhA -> ME
- LAU, TEI, FOI, BY -> BY
- SA, GUhA, BIhE, CEhE, PEhE, NUhI, NUhU, MAhO, NAhU, PEhO, KUhE, FUhA, MOhE -> /
Allow after any selbri. It's a common mistake, it has an obvious meaning, it doesn't cause parsing difficulties.