On 14 July 2012 05:36, la gleki <
gleki.is...@gmail.com> wrote:
>
>
> On May 29, 7:43 am, Jacob Errington <
nicty...@gmail.com> wrote:
>> Yes, there is. The gismu deep structure is wrong:
>> lo ka bavbavlamdei cu ka ce'u djedi gi'e lamji ce'u noi balvi *lo balvi be
>> ce'u*
>> where the added on "lo balvi" is the result of tacking on the extra -bav-.
>> So "a day that is adjacent to an event that is in the future of the future
>> of an event" i.e. BAD.
>>
>> In fact, the GDS for bavlamdei is really messy:
>> lo ka bavlamdei cu ka ce'u djedi gi'e lamji ce'u noi balvi ce'u
>>
>> You might as well just use {ba za re lo djedi}
>>
>> My ramblings:
>> {lo se detri be li za'u} is an interesting solution, where the number of
>> "tomorrows" can be indicated after the za'u:
>> {lo se detri be li za'ure} -> "the day after tomorrow" and {lo se detri be
>> li za'uci} "the day after that".
>> If {za'u} had a lujvo, which is sadly does not, I'd propose: {za'u zei
>> seldetri} = x1 is the event occurring on date x2 days after x3 at location
>> x4 by caldendar x5.
>> Unfortunately {za'u zei seldetri} suffers from karcykla syndrome: is it
>> days, months, weeks, years? *shrug* we can't know
> Shouldn't we add {pi'e} to point to week/month/year?
have no idea what you mean by "add {pi'e}". We can't add it to the