Advanced Place Structure Mangling
We all know that be works on the selbri-level, allowing us to inject sumti directly into the selbri. We also know that the primary use for this is in description sumti, where we can select an x1 based on definite (i.e. non-zo'e) values in other places. However, being done on a selbri-level, what's really happening with be is that we're creating "new" selbri each time we use it.
The Real Purpose of be
Consider lo klama be la .bastyn.: we create a new selbri with the definition "x1 goes to Boston ..." in order to select the x1 and get a referent and whatever. (The actual way articles work and possible issues with that is also outside the scope of this e-mail.) What's really interesting, however, isn't description sumti. (In fact, that's pretty boring.) What's really interesting is what is in the "..." of my previous definition of klama be la .bastyn.. The full definition is "x1 goes to Boston from x2 via x3 by means x4."
Notice how the x3 place of klama has become the x2 place of klama be ko'a. Indeed, injecting a sumti has the effect of moving all the later places forward. What repercussions does this have on top-level bridi?
Consider dunda, "x1 gives x2 to x3." Given the above proof of place promotion, dunda be ko'a should mean "x1 gives ko'a to x2." Considering that ko'a broda ko'e equals by definition ko'a ko'e broda, mi do dunda be ko'a means "I give you ko'a," mirroring the English structure! By using be and the ability to move places into the bridi-head, we can create pseudo-"1 3 2" argument order at the cost of one syllable. The actual SE conversion required to achieve the true structure is se te se which is three syllables long and requires forethought. In a way, this selbri-level manipulation can almost be thought of as "afterthought SE conversion."
Taking the Idea Further
We can create even wonkier place structures by using be FA. Basic se can be achieved using ko'a broda be fa ko'e, e.g. lo nu lo pampe'o cu darno cu badri be fa mi "My lover being far is what saddens me." The real complexity of the transformation comes from moving bridi-tail sumti into the bridi-head: .i mi lo plise lo barda cu vecnu be fi do "I an apple for a lot sold to you." (Mirroring the Lojban with English leads to ugliness, but I'll try to keep it up so long as it's (semi-)understandable.) The argument order in that case is 1-2-4-3, with a corresponding SE conversion of te ve te.
From this, I concluded the generality that when all sumti are moved to the bridi head, the effect of broda be fa xi ny ko'a is to send the x_n place to the end of the place structure. In a way, this can be thought of as "remote" FA, as it allows us to perform a FA operation on the selbri level, outside the formal place structure.
An example, with corresponding argument order and SE conversion: .i lo tcadu zdani lo nurma zdani cu klama be fa mi "To the city house from the country house go I." 2-3-1 se te
Before continuing, we must recall that se is applied before be, as lo se broda be ko'a has ko'a in broda1.
Combining this type of be operation with ordinary SE conversions, we can produce extremely cryptically ordered selbri: .i lo ni se pluka lo ka pinxe lo ckafi kei kei do te zmadu be mi) "In the amount of enjoying drinking coffee, you are greater than me." 3-1-2 te se What is truly shocking about this is that we see a te-conversion, but the selbri becomes reduced into a binary predicate, i.e. a predicate with two argument slots, because of the be ko'a.
ke..ke'e-Brackets as Selbri Boxes
It is (currently) ungrammatical to use be twice on the same selbri. (Not that I'd necessarily want it to be.) Indeed we have bei for that. But, we can think of the following as "forethought bei."
Using ke to create a bracket, we can box one selbri inside another, ke broda [ke'e] has the broda selbri trapped inside the ke..ke'e selbri. The ke..ke'e selbri has the same structure as the inner selbri, and this is what we can exploit to avoid using bei. Formally, ke broda ke'e be ko'a is equivalent to broda be ko'a, but if broda already has linkargs, i.e. injected sumti with be, we can't move the outer be inside.
Consider klama be la .bastyn., "x1 goes to Boston from x2 ..." We can't attach another be to this selbri because it would be ungrammatical, but we can box it inside ke..ke'e and make use of the place structure transparency outlined in the previous paragraph and then use be: ke klama be la bastyn [be'o] be la montre'al, "x1 goes to Boston from Montreal via x2 in vehicle x3. In cases where the first sumti does not end in a selbri, the use of the second be will cause (a lot) of elision.
Although "forethought" bei might appear utterly useless at first, it can be used to avoid using multiple FA. Suppose we want to specify the x3 then the x2 with injected sumti. Normally, we need to use broda be fi ko'a bei fe ko'e because that's the way it works when you only have one selbri, but when we use ke..ke'e as a selbri "box", we get to cheat, by considering that the inner selbri has its own place structure: .i mi fi lo karce cu ke klama be fi la .montre'al. [be'o] [ke'e] be la .bastyn. First, we consider the inner selbri klama be fi la .montre'al. as having the place structure "x1 goes to x2 from Montreal via x3 in x4." Then, we box that selbri inside the ke..ke'e brackets, and use be again, to fill the x2, i.e. the destination, with la .bastyn.. This yields the place structure "x1 goes to Boston from Montreal via x2 in x3." Finally, when we do formal place filling, we use fi to specify the vehicle, in this case lo karce "I, in the car, go from Montreal to Boston."
Although using ke..ke'e by itself to create a place structure-transparent "box" is pretty interesting, we can create even what I consider to be the most advanced structure changes by using a combination of all the tools outlined above: SE, ke..ke'e-boxing, be, and bei.