{lo'i} as a Q-kau solution?

["And Rosta" <a.rosta@ntlworld.com>]

During our first major Q-kau discussions, around 1997 or so 
(I estimate, without checking the archive) -- or perhaps this
was more recently during the What I Have For Dinner thread?
-- I remember I suggested things like:

mi djuno lo'i klama = "I know who goes"

[= For every property of lo'i klama I know that it is
a property of lo'i klama --?]

As I recall, this was quickly dismissed, and we moved on
to other solutions, but I'm now wondering why it was
dismissed. If a set's only properties are the identity of
its members, then (for example) knowing a set must mean
knowing which things are (and aren't) members of the set.

An unthorough check through all nonmain-bridi Q-kau
contexts (djuno/know, kucli/wonder, change, depend on, 
discover), suggests to me that only x3 of frica (ka
ce'u prami ma kau) is not readily amenable to this solution,
but this doesn't seem an intractable problem.

As for main-bridi Q-kau (Q-ever kau), I would now like to
concede to Jorge that these are genuine Q-kau constructions,
but (I think) all Q-kau reduce to {xu kau}, and AFAICS
(myopically) {xu kau} = "whether or not" = U-connective.

I can't believe the Q-kau problem can be solved so gordianly,
so please point out the problems that I am failing to
recognize...

--And.