From Http://Members.Aol.Com/Lassailly/Tunuframe.Html wrote:
>=20
> Dans un courrier dat=E9 du 28/05/99 13:10:17 , Josh a =E9crit :
> > Exactly right! That's why I usually refer to my objects by the gene=
ric
> > 1st, 2nd and 3rd rather than accusative, dative, "jurative" (the adv=
antage
> > of these being that the names are more memorable -- good for teachin=
g the
> > language, as opposed to theorizing about it); they don't have univer=
sal
> > meanings, but depend on the definition of the particular word, or in=
most
> > cases that of its class.
> planning roles into
> standard maps as you do is necessary up to 2 or 3 roles. This implies t=
hat
> the mapping of transitive and ditransitive voices are more precisely
> designed. Even a very simple language like Indonesian makes a differenc=
e
> between final transitive (me-) processive transitive (me-kan), ditransi=
tive
> transformative (memper-kan) or address (mem-i) because "to do something=
" or
> "to have somebody do something" are not the same map as "to make someth=
ing
> into something else" or "to apply something on something else". It's ve=
ry
> stupid to have different prepositions for the core roles. Some are univ=
ersal,
> some others are not. It would be better to have one preposition specifi=
c to
> 2nd object for transformative, one for address, etc. Japanese "ni" and
> Indonesian "pada" are good for that.
Lojban makes them all arbitrary, but well-defined:
http://www.animal.helsinki.fi/lojban/gismu.html