The lerfu shifts (BY1) consist of these cmavo: ga'e, ge'o, je'o, jo'o, lo'a, na'a, ru'o, se'e, to'a.

Usage observations

For past usage, I am searching through the corpus I have collected of 900 kilobytes of pure-Lojban text. The corpus includes all texts published at lojban.org/files/texts, many large texts from the Wiki, IRC logs, and texts from the CVS server such as Alice.

ga'e, to'a (case shifts)

ga'e is only used correctly in algebra, to mark variables named with capital letters. The author assumed that the shift would apply across multiple lerfu strings.

to'a is never used correctly.

lapoi pelxu ku'o trajynobli contains the sentence ".itu'e ga'e ca cpedu fi do to'a". ga'e and to'a here both act as pro-sumti, which was probably not intended. Here, ga'e and to'a were probably intended to "capitalize" (emphasize) the words between them, but as lerfu modifiers they cannot modify the emphasis of words.

ge'o, je'o, jo'o, lo'a, ru'o (alphabet shifts)

None of these are used anywhere in the corpus, except that the utterance "zo ru'o" appeared on IRC in response to a line of Russian text.

zai has not been assigned to this section, but it has a similar function to the above cmavo. It is also not used anywhere in the corpus.

na'a (cancel shifts)

This word is not used in the corpus, though nau was used in the algebra text where na'a was probably intended.

se'e (character code)

This word is not used in the corpus.

Proposed definitions

ga'e

Converts future letterals to uppercase. The change applies until it is shifted back with to'a or cancelled with na'a.

na'a

Cancels all shifts (font, case, etc.) currently applied to letterals. Any shifts that occur earlier in the text do not affect letterals from this point on.

se'e

Convert the next sequence of digits to a character code in ASCII, Unicode, or some other agreed-upon character set. The code includes all digits until the next non-digit, the end of the letteral sequence, or na'a.

to'a

Converts future letterals to lowercase.

ge'o

Converts future letterals to the Greek alphabet.

je'o

Converts future letterals to the Hebrew alphabet.

jo'o

Converts future letterals to the Arabic alphabet.

lo'a

Converts future letterals to the Lojban (Roman) alphabet.

ru'o

Converts future letterals to the Russian (Cyrillic) alphabet.

Proposed keywords

ga'e: uppercase shift, uppercase
na'a: cancel shifts
se'e: character code
to'a: lowercase shift, lowercase
ge'o: Greek shift
je'o: Hebrew shift
jo'o: Arabic shift
lo'a: Lojban shift, Roman shift
ru'o: Russian shift, Cyrillic shift

Changes

Clarification of scope

The scope of a letteral shift needs to be defined. I will elaborate on Arnt's specification in BPFK Section: lerfu Forming cmavo, also following the "Microsoft Word model" specified at Interpretive conventions for lerfu formatting cmavo.

A letteral shift lasts until another shift of the same type replaces it, or it is cancelled by na'a.

(The sole usage of ga'e assumed that it would not end at the end of a lerfu string.)

It is not so far specified where a se'e construct should end; I propose that it should be able to be terminated with na'a, because na'a terminates other sorts of shifts.

Omission of unused cmavo

Given that Lojban does not seem to be intended for holding multilingual spelling bees, and that a dictionary containing many unused cmavo with bizarre functions could confuse learners of the language, the BPFK does not recommend to include the alphabet shifts (to'a, ge'o, je'o, jo'o, lo'a, ru'o, zai) in learning materials intended even for advanced learners.

Impact

The clarifications made to the scope of lerfu shifts give a consistent model of how shifts should be applied, and do not invalidate any known usage.

I believe that my scope clarifications are consistent with those in BPFK Section: lerfu Forming cmavo, even though that page says otherwise.

Given the lack of usage of alphabet shifts, omitting the unused alphabet shifts from learning materials should not have any significant impact on the language.