It sucks that you only get fai when using jai. There are many cases where it would be cool to get a fai'd non-x1 place. The fix {SEjaiSE} is ugly and wordy and Lojban deserves better.
Let's say we want to say "I allow you to kiss me".
There are two ways that require nothing new to be invented:
(1) mi curmi lo nu do mi cinba
     "I allow that you kiss me."

(2a) mi se jai se curmi do fai lo nu cinba mi
(2b) do jai se curmi mi fai lo nu cinba mi

Sentence (1) is of course fine, but suffers from one small drawback: It requires forethought. If you already said {mi curmi tu'a do}, you are screwed and will be the laughing stock of your friends (zo'o).
Sentences (2a) and (2b) are similar, but (2a) is much uglier. (2b) I would consider quite normal and beautiful, but here we had to rearrange the sumti to make it work, again requiring a certain amount of forethought.

What I propose:
Allow fai to refer to tu'a places when no jai is present:
(3) mi curmi tu'a do fai lo nu cinba mi
     "I allow you to kiss me."

If people complain about this, then at least invent a FA that does this job. It could then be used even when a jai is also present. I propose fau'a in that case. However, I don't like inventing new cmavo all the time. In my opinion, it's better to extend the use of existing cmavo.

I came across this image and I'm really intrigued. What could this be?

Here I want to list all the cmavo rafsi that I and tsani have used. My motivation for these rafsi is that it adds yet another type of language to the lojban foreign-language-support: Polysynthetic languages.
Using the existing cmavo plus the ones presented in this list, I hope for Lojban to be able to express any bridi within a single lujvo, much like polysynthetic languages would do it. Here's the list. I'll be updating it once more rafsi are found.

FA rafsi
fa    faf
fe    fef  
fi    fif
fo    fof
fai    fox

JAI
jai    jax

BY (i know that one or two of these aren't free, I will check later)
by    boz
cy    coz
dy    doz  
fy    foz
gy    goz
jy    joz
ky    koz
ly     loz
my    moz
ny    noz
py    poz
ry    roz  
sy    soz
ty    toz
vy    voz
xy    xoz
zy    zoz

Other cmavo
ma    ma'y
mo    mo'e
ko    kov
po'o    pov
kau    kaz

I propose to allow any VUhU to appear anywhere a JOI can.
This makes sense because VUhU are (usually binary) operators just like JOI and using them in other places than mekso makes sense.

Advantages:
# More flexibility
# Allowing mekso to be used in more places = more utility of a huge part of cmavo space

Disadvantes:
# Pretty much none. It makes things grammatical that weren't before, but it does not break any usage.

Definitions must be found for all the use cases, e.g.:
SUMTI VUhU SUMTI, SELBRI VUhU SELBRI etc

If VUhU *moves* to JOI, JOI has to get all the mekso definitions as well to make things like (1) grammatical:
(1) li pe'o joi paboi paboi du li re

So it might be easier to just give VUhU more rights and allowing it in places that JOI can go, but not vice versa.

Examples

(2) lo mu bakni su'i lo mu kanba cu zilkancu li pano lo danlu
"5 cows + 5 goats amount to 10 in animals"

(3) lo mi lanzu vu'u ba bo lo mensi noi cliva bilga cu zvati lo nunsla
"My family minus-later my sister, who had to leave, were at the party."

This removes some of the awkwardness of mekso and makes it more intuitive and less verbose. I opine that the above example sentences show a useful application of what this proposal entails.

This proposal is not very detailed yet, but I can flesh it out, should there be desire for constructive discussion.

One of Lojban's design goals is audio-visual isomorphism, one-to-one mapping of speech and writing. I think this goal is a very important one and worth to be aspired. Lojban has this feature to 95%, but why not 100%? There seems to be no good excuse. It's just lazines. The way I see it is this: For speech streams and streams of letters to be 100%ly phonemic and isomorphic, there has to be one letter per sound and vice-versa. Therefore, having i/u behave like approximants in some cases and as vowels in others as well as allowing glottal stops, which are part of the speech stream, to be elided from writing seems horribly inconsistent to me and is an unnecessary concession. What follows from this is the only reasonable solution, and that is to have a writing system that maps to speech 100% in the following way:

1. Nothing that is pronounced must be omitted in writing. This includes glottal stops as well as stress(!)
2. Spaces must not matter. Speech streams contain no whitespaces, they segment because of stress and clusters. Therefore, every written text must be unambiguously parsable with and WITHOUT spaces. This means that any text must be written in a way that doesn't require the spaces for it to segment properly. Spaces are TRULY optional (unlike now, where spaces are optional only as long as you watch out for bullshit exceptional stress marking). Since now spaces are truly optional, they are only there to facilitate reading the text, not to enable it.
3. Give every sound a SINGLE phonemic letter. iu becomes ĭu, ua becomes ŭa. This also makes it less confusing for people that never understood why denpa bu is not needed before approximants.
   

Here's a sample text using this system:


                                               lo sònci be bu'u lo bùmru  

ni'o lo dàplu no'u la. gont. zi'e noi sèlpau pa cmàna ku po'o noi lo galraìpau be ke'a cu gàlmau lo vilbì'e se bikydà'i berstunyxàmsi lo mìnli cu zìlsau lo ka gùgde lo màfcre .i vei na'e bo so'u lo nànmu xàbju be la. gont. cu clìva lo za'u tcàdu pe lo gàltu ma'arbì'i ge'u joi lo jàrki je mànku zbàni te zu'e lo nu go nai sèlfu lo tùrni be lo dàplu tùtra bu'u lo ri tcàdu lo ka termàkfa gi'o nai màfcre gi mu'i lo zancapsèlfri cu lìtru te ka'a pa dàplu se ka'a lo dràta gi'e pìlno lo màkfa bu'u pi ro lo tùtra pe la tedyxàmsi .i cmìma ro lo bi'u nai prènu —sei so'o da cu jìnvi— fa de poi lo dràta cu te cmène fi ke'a fe zo cpinìsu zi'e po'u lo bàlrai .i je'u tràji lo ni lìtru kei gi'e zanselfù'a lo nu bìnxo lo dràkytru kei gi'e ba za bo se cmicù'a bìnxo lo ralmàfcre .i ba'e fu'e la me la. ged. selsà'a fu'o .e so'i dràta selsà'a cu te lìsri lo ra nunzù'e [...]


In addition, here are the UI that use approximants:
ŭa, ŭe, ŭi, ŭo, ŭu
ĭa, ĭe, ĭi, ĭo, ĭu

The diphthongs:
All of [ɪ j i] and [ʊ w u] are legal realizations of the final V in a {.VV}, which suggest there being an additional phoneme for each of those groups. It cannot be analyzed as a consonant in terms of morphology. A VV counts as a single vowel, just the way it should be for a diphthong. The ĭ represents the consonantal approximant, which, in Lojban, is only important within a CV context, e.g. {ĭu}.

ụ ị

Will rewrite this article soon