doi tsani
I see things a bit different. I believe that kakne2 and djica2 should
be abstracts
({ka} and {du'u}), but for their intensionality. gleki2, on the other
hand, is as
concrete an event as it can be, not an abstract property or
proposition. Consider
{mi gleki so'i lo nu mi klama lo zdani be mi}
Actually, quantifying over events like that has various problems, namely that the zo'e inside it are magically changing. In Lojban, every event is unique, but as humans we roll them together.
In reality, what's happening when you quantify over an event like that, you want to make a whole bunch of slightly different events: the event of going home on Monday, and the one of Tuesday, and the one on Wednesday, etc. Each of these events in unique.
It seems like a predicate logic sentence would be best to describe this, and ultimately moves the quantifier scope higher: {.i so'i da zo'u mi gleki lo ka klama lo zdani ca da} "For many X, I am happy to go home at X." This ultimately creates many properties of going home, for each of which it is asserted that the speaker is happy about that.
Also, your
{mi gleki lo ka se li'i do citka lo plise} instead of
{mi gleki lo nu do citka lo plise}
is like
{mi klama lo stuzi be lo zdani be mi} instead of
{mi klama lo zdani be mi}
or
"I saw an image of the sun setting" instead of
"I saw the sun setting",
This extreme typing of sumti places just creates hindrances in
_expression_ without
adding anything to the speaker's or listener's understanding of the
world. I already
know that it is my experience of an event that can bring me happiness about it,
not somebody else's, and that what determines whether someone is tall or short
is his/her body, not his/her friendliness.
As I explained, the general benefit of making places infinitive-places with {ka} is to make jvajvo more commonplace. As it stands, we frequently drop "repeated" places that it would be nice to still have. Reconsider my examples of {ctidji}. Using current nu-based djica, there is no way to get both meanings in separate lujvo. Making the lujvo always drops the other one. If we define ctidji with a place merger, we can't recover ctidji without that merger.
ctidji = x1 djica lo nu x1 citka x2 ...
or
ctidji = x1 djica lo nu x2 citka x3 ... ?
Defining one makes the other undefinable, whereas with infinitives, we just have to throw in -fri- to get the other meaning:
ctidji = x1 djica lo ka ce'u citka x2 ...
and
ctifridji = x1 djica lo ka ce'u lifri lo li'i x2 citka x3 ...
I agree, it creates some boilerplate if you want to say something like "I want you to eat an apple," but there're ways to circumvent this. In fact, I frequently use this tanru trick to avoid using {lo nu} and {lo ka}: {.i mi djica co li'i do broda}. This has the unfortunate disadvantage of making it impossible to assign djica3 (or explicitly assign djica2, although tanru inference tells us that {lo ka se li'i do broda} is djica 2).
Another way to "cheat the system" would be to use sei, assuming sei functions as a bridi relative clause (which I believe is the most sane interpretation of sei and as I see it, breaks no usage whatsoever): {.i sei mi djica do broda [da'i]}. We might need to include {da'i} in some cases, where subjunctives happen implicitly in the nested bridi. In the case of djica2, we usually assume that the djica2 is subjunctive, i.e. is true hypothetically, but not in reality. My example with sei, however, moves the nested bridi onto the top level, pushing the djica-bridi into a relative clause. Strictly speaking, the nested bridi becomes claimed, which is undesirable in the case of djica.
(The choice of li'i and lifri in general for more flexibility in the rigid infinitive system requires that the actual association between the lifri1 and lifri2 possibly be extremely vague, which isn't exactly a problem.)
As for your comparison between {gleki lo ka se li'i broda} and {klama lo stuzi be ko'a} isn't entirely rational. The "speech cost" of {ka se li'i} versus the introduction of a whole other selbri and linked arguments is completely different. {ka se li'i} versus {nu} is only 3 syllables (short syllables with no consonant clusters) longer. (At most four, if you include the extra kei.) Also, I find it much easier to understand a compound abstractor like this than complex linked arguments.
Also, stuzi1 and stuzi2 are arguably *the same thing*, which can be troubling. Let's take a look at another selbri for comparison. Although it's experimental, {ditcu} is the best example of this: "x1 is the time-duration of x2." It turns out that both the x1 and x2 are events, because in Lojban, there isn't actually a system for representing quantities, so we use objects (or events) as proxies for their relevant properties. That being said, ditcu simply states an equivalency in the event-bridis' ze'a-tag: {ko'a ko'e go ditcu gi dunli lo ka makau jai ze'a fasnu fai ce'u}. Now to get back to stuzi, does stuzi simply state an equivalency in the objects' bu'u-tag ? {ko'a ko'e go stuzi gi dunli lo ka makau jai bu'u me ko'a} ? I'm not entirely sure. It doesn't really look like it, because that means I get to say things like {.i lo zdani be mi ca stuzi mi}, which shouldn't really make sense. However, that being said, *what* kinds of things can we even *put* in stuzi1 ?
Now, if stuzi1 and stuzi2 are argubly the same thing, then {mi klama lo stuzi be ko'a} and {mi klama ko'a} are also arguably the same thing. However, under my infinitives system, events are not interchangeable with ka-abstractions, and there is no way to define an equivalency as I have in the case of stuzi.
.i mi'e la tsani mu'o