Weird case marking patterns

I just borrowed a somewhat new book on linguistic
typology

Author 	 Song, Jae Jung
Title 	Linguistic typology : morphology and syntax / Jae Jung Song
Publication 	Harlow : Pearson Education, 2001
Material Information 	406 s. : ill. ;
Series 	Longman linguistics library,
ISBN 	0582312213 (pbk.) : £16.99
ISBN 	0582312205 : £40.00 : CIP entry (Aug.)

and true to my nature I immediately read the chapter on
case marking.

In the section about mixed marking systems it says,
that there are languages which have an ergative-
absolutive marking system on NPs and nominative-
accusative marking on verbs while there are none
that have it the other way around (i.e. no language
with nom-acc NPs and erg-abs verbs!), but it doesn't
name any language which has this marking system.

It also mentions the _Nominal Hierarchy_:

^   1st person, 2nd person
|   3d person
|   personal name/kin term
|   human
|   animate
|   inanimate

In languages with split ergativity categories towards
the top of the hierarchy are most likely to have nominative-
accusative case marking while items towards the bottom
are most likely to have ergative-absolutive case marking.
Again there are no known languages that violate the hierarchy,
i.e. having erg-abs on 1st/2nd person pronouns and nom-acc
on inanimates.  Languages differ WRT where in the hierarchy
they draw the border, but they don't wiolate the hierarchy.

Now this has some implications for my (perpetual) thinking on
the Sohlob case system.  In my most recent delineation animate
agents take the ergative case, while inanimate agents take the
instrumental case -- i.e. inanimates are not capable of true
agentness.  This would give me a nice division of labor between
the _-l_ case -- which in slightly different functions
have been present in the precursors of Sohlob since the late
seventies (1) -- and the _-r_ case, the _-l_ case being
ergative and the _-r_ case instrumental.  They are supposed
to be historically-phonologically related in that the pre-
Sohlob language had *r_j and *r rather than /l/ and /r/.

Now imagine combining this with the Nominal Hierarchy,
so that 1st person and 2nd person take nom-acc marking,
inanimates take "instrumental-absolutive" marking and
the categories inbetween take erg-abs marking!  Of
course both the nominative and the absolutive are to
be "marked" with a zero morph!

Now I wonder if such a doubly complicated case marking
system might be attested in any natlang, or if it is
too weird?  Of course I also contemplate combining this
NP marking system with consistent nom-acc marking on the
verb, which is perhaps too unrealistic, or perhaps even
with no marking at all on verbs for inanimates but nom-acc
for categories higher in the hierarchy, which potentially
is super-unrealistic...

--
/BP 8^)>
--
Benct Philip Jonsson -- melroch at melroch dot se

    "Maybe" is a strange word.  When mum or dad says it
    it means "yes", but when my big brothers say it it
    means "no"!

                            (Philip Jonsson jr, age 7)