The talk is of {voi} and {soi}. I consider both of them dispensable with their current meaning. {voi} is so rarely used, and it has no non-restrictive counterpart. I'd much rather like to have a UI that expresses non-veridicality (a discursive "I describe"), than have {voi} and {le} engage in such things. Then {noi UI} and {poi UI} give the two descriptive NOI, and {lo UI} gives non-veridical {le}. I most likely don't have to convince many why {soi} wouldn't be missed. It has already fallen out of use, and the only reason one could argue for keeping it is that reassigning it would break usage, but this I consider a weak argument; is Lojban supposed to keep a unused cmavo just because some old texts use it? Lojban (hopefully) has most of its time still before it. Now is the best time to make changes, not in 50 years when it's all too late. The language is still small right now. There are no fluent speakers, a tiny corpus, and barely any real-world use (how could there be, with nobody to speak it with, and with such a limited lexicon?).
So what would I do with those two cmavo? First of all, I'm not sure, not comitted to the following ideas. They are probably better than what is currently done with the cmavo, but that doesn't make them good automatically!
The first topic is that of "as": Check out this mailing list discussion to get an idea. If you read through the thread, you'll see that one suggestion is to use {poi} to restrict an individual to a certain capacity:
(1) Mi nelci zo gumri poi cmene .i ku'i na nelci G poi gismu
"I like "gumri" as a name, but I don't like it as a gismu."
(2) Mi sinma do poi pendo .i ku'i do mabla gunka
"I respect you as a friend, but you are a sucky worker."
(don't mind the capital letters, I'm just playing around a little)
Some people dislike using {poi} like that. And it's understandable, since {poi}'s primary purpose is to restrict from a set to individuals (but if you read on, you'll see why it might not be a very different idea from that). There is also the argument that {do} only has a single referent to begin with, so restricting it achieves nothing, it's still {do}. Okay, but why can we say so easily comprehend things like "I dedicate this book to the little boy he once was" or "the you of the past"? Evidently, we think of the "you of the past" as a different thing from "the you of right now". And I don't think that's weird: I like to view objects not as a single entity that is constant throughout time and space, but as an amalgamation or series of many many small slices of that object's existence. This makes sense to me, because the person that is writing this article right now is definitely not identical to the one that first discovered Lojban some years ago, although they both go by the same name (well, actually I changed my name...). And time isn't the only factor. I clearly behave differently in different situations and environments. I might be confident in one situation but completely lost in another. Am I, as a whole, confident or not then? Both? None? na'i? Probably na'i. It simply depends on the where and when. That's why I think it's not bizarre to want to refer to just slices of an objects as opposed to the whole thing at once. I consider words like {mi} to be agnostic about slice-ness. They are much like {lo broda} in that they can refer to anything ranging from a mere glimpse of Broda to a large group of Broda, from Broda in a single place to Broda ever and anywhere. Each one of those slices is still Broda, or {mi} or {do} or {zo gumri}. If I touch you here and now, I can safely say I have touched {do}, but have I touched the person (that specific slice) you were 10 years ago? I don't think so. Still, both are {do}, a huge conglomerate of seperate {do} pieces, so huge that the whole can never be involved in a single situation limited to a point in time-space. But, and this is also interesting, it can leap through time and space. I can now like the you of when we first met, although I now can't stand you anymore. I can import past slices of you into a present context.
So with the above in mind, I am confident that there is sense in wanting to restrict to slices. If {poi} is being rejected because it's supposed to be solely about restricting the domain of quantification, then I can accept that. So I would like to use {voi} or {soi} in place of it in sentences (1) and (2). I assume {voi} is the better candidate, since it's already in NOI and the use I could imagine for {soi} would be in SEI, so it's overall less of a change in terms of grammar than if done the other way around.
So then we'd get:
(3) do voi mi pa re'u penmi
"you as I first met you"
(4) mi sinma ge do voi pendo gi nai do voi jatna
"I respect you as a friend, but not as a leader."
The other topic, which is probably less controversial, is that of bridi relative clauses. If you followed my venturing into Gua\spi-Land, you might remember that it does a lot with bridi relative clauses. I'm sure Gua\spi has been an inspiration for those few, courageous, Lojbanists who went ahead and gave Gua\spi a closer look. Not only did it probably inspire the use of {sei}+{ke'a} as a bridi relative clause in Lojban, it also probably helped the {ka}-volution thanks to its very ce'u-esque gismu definitions. Now the thing with {sei} is that it has seen a lot of use, unlike {voi}. And {sei} is extremely useful as the metalinguistic parenthetical that it is. It's flexible, it is easy to use because it doesn't have any restrictive rules that limit its applicability. It lets the Lojbanist relax, it lets them be vague. It lets them relate the main bridi to the sei clause in what ever way they want, or in no way at all, if desired. Something that is great about SEI is that it seems to be the only thing (apart from {to ... toi}) that allows one to escape scope. This is invaluable. It's often overlooked, but Lojban is hard to speak correctly once you introduce a couple bridi operators. You have to be extremely careful to place them in the right order, or else you end up saying something very different from what you intended. This means you have to completely reconsider everything you say, and, essentially, have to forget about English. With bridi operator rules strictly followed, you cannot copy every word order anymore. You have to realize that Lojban has its own word order, which depends solely on scope. Complex Lojban spoken properly is much more like Japanese than English. Everything is forethought, because that's the only option; you can never go back to fix your scope, it has to be obvious to you from the get-go. That's why SEI is so damn powerful, at least potentially. It simply ignores scope, or in other words, in puts itself right before the earliest prenex. This makes SEI unique. It's a safety net, among other things. But again, it's vague. la tsani liked the idea of using {ke'a} in the SEI, where {ke'a} would refer to the main bridi, but it doesn't seem to have caught on. I'm not sure he is still in favor of using {ke'a} with {sei}, but he probably still likes the idea of an actual bridi relative clause. Here is an example, to make it more concrete:
(3) mi cadzu lo canxaskoi sei ke'a li pa cacra
"I walk on the sand beach for one hour."
lit. "I walk on the sand beach, one hour being the amount of hours that it [= my walking on the sand beach] lasts."
A nice thing to have. The problem is that it breaks all those vague usages of {sei}, which I hope I was able to show are desirable to have. Sure, the {ke'a} could be argued to behave just like it does in a NOI; vague and implicit, and able to appear in the darkest, most nested corner of the NOI. But it would still mean that a ke'a has to be in it. I think it's fine to keep {sei} the way it is, as a sibling of {to}. Just comments, without any restrictions.
But, because I do think a pure bridi relative clause as portrayed in (3) would be a good thing to have, I would propose {soi} to be moved to SEI, and given the function that {sei} took in (3), so that:
(4) mi sipna soi do ke'a zgana
"I sleep, which you observe."
Note that this new {soi} would also be a scope leaper, like {sei}:
(5a) su'o roi ku mi ze'a lo cacra be li za'u pa mu cu sipna soi ui no da djuno
"Sometimes I sleep for over fifteen hours, which fortunately nobody knows."
Pay special attention to the scope of {no da} in relation to the other two bridi operators {su'o roi ku} and {ze'a}. It can be observed that {no da}, despite being to the right of them, has scope over them. This fact deserves to be appreciated fully. Sentence (5a) can be rewritten thus:
(5b) ui no da djuno lo du'u su'o roi ku mi ze'a lo cacra be li za'u pa no cu sipna
"Fortunately, nobody knows that sometimes, I sleep for over fifteen hours."
So these might be worthy candidate meanings for {voi} and {soi}. I hope my presentation was able to convey the possibilities of and motivations for such a change. I invite you to give it some thought. Have a nice day!