Re: [lojban] Re: Questions about the Magic Words

[Raphaël Poss <grey.havens@online.fr>]

Le 9 nov. 04, à 20:17, Robin Lee Powell a écrit :

> On Tue, Nov 09, 2004 at 01:40:25PM +0100, Rapha?l Poss wrote:
> > I am now reading the Magic Words wiki page, and it actually
> > challenges what I thought I previously knew about the lojban
> > grammar.
>
> I bet.  Please note that xorxes and I are discussing a completely
> different version, as well.

Ah. Is it on the wiki ?

> I would like to know how hard you think it would be to memorize the
> rules as I've listed them.

*I* for one think the rules themselves are sufficiently simple to be 
learned easily, *but* :

- the phrasing is poor and/or too complex,

- it raises implicitely a lot of questions about special cases, which 
should be provided for illustration beneath the rules.

About the phrasing, I would propose to keep the formal rules with any 
level of precision needed for them to be actually formal rules, but 
accompany each one of 2 "summaries" : one simplified grouping rule and 
one human-readable summary. See at the bottom of my message for an 
example.


> > I would very much appreciate if someone could enlighten me on the
> > following points :
> >
> > 1. "zo y bu == (zo bu) [one y ignored]" or "zo y bu == (zo (y
> > bu))" ?
>
> The former.  If .y. wasn't ignored, the latter would be an error.

Then for any of :

  zoi y bu ... bu
  zoi y bu

Rule 1 states "ignore Y except before BU" : since no quoting has been 
done yet, Y is not ignored at this point,

Rule 2 and 3 do the quoting, saying that Y is not taken into account.

After rule 3, what remains of Y ? Is it embedded as some kind of 
meta-data in the quote, or is it dropped afterwards, or what ?

It is tempting to rephrase rule 1 as:

  - Y is always ignored after ZO or ZOI
  - Y is not ignored before BU
  - Y is ignored in all other cases

But then we have a problem:

  zoi zoi ... zoi y bu

Since the quoting is not done yet (at this point), Y is forcibly 
ignored, changing the meaning of the phrase.

I'm confused. Also, see point 3 below.

> > 2. is "broda lo'u brode fa'o" grammatical ? (can we assume that
> > fa'o closes all elidable terminators, even le'u ?)
>
> That's a parser design issue, really, but as it stands, no.  Bear in
> mind that le'u is *NOT* an elidable terminator, in that it is not
> elidable.

Therefore, since fa'o closes the input before the lo'u quoting is done, 
the construct becomes ungrammatical.

Now, wait, I can see that now fa'o can be quoted with lo'u (rule 4, has 
it changed ?)

So "broda lo'u brode fa'o le'u" is now grammatical.


> > 3. what remains after "zo y bu si" ?
>
> An unfilled zo.

Given the current rule 1, I would say "zo y". (Y was not ignored since 
BU happened just after).

> > 4. what remains after "zoi zoi zo zoi sa zo" ? (I ask because I
> > notice sa is processed before zoi)
>
> The stuff inside zoi is not processed as Lojban, and hence has no
> selma'o, so nothing remains.

But zoi was not processed at the point SA is processed (but since the 
wiki page has changed). Now it's ok.

> > 6. I notice that fa'o is processed before zoi : what remains after
> > "broda zoi zoi fa'o" ? (is the quote automatically closed, or does
> > the sumti become ungrammatical ?)
>
> Neither; the fa'o is quoted by being inside a zoi construction.
>
> Clearly, zoi should be closer to the top of the list.

Now that it is, things have become saner. Thanks :)

Now, more questions !

Point 8.

  what happens if the word after SA does not appear before at all 
(never since the beginning of the text) ? Is the erasing performed, or 
ignored ? What remains ?

  Especially, what remains after :

  - broda sa su

  - broda sa sa

  - broda su brode sa su

  - zo sa broda sa sa

  Ignoring it would allow cancelling an erasure in progress. For 
instance, if I start to utter "broda sa ..." and I want to cancel the 
erasure, I could continue with "... sa si si" (since no sa could have 
remained from before that way)


Point 9.

  What remains after :

    pa zoi pa ... pa sa pa

  (or, is the ZOI delimiter valid as a limit for "sa" ?)


Point 10.

  My current knowledge of lojban tells me that :

    la'e lu zo zo li'u

  means "the word 'zo'".

  Since nothing prevents lo'u...le'u from quoting correct text, I 
assume that "la'e lo'u zo zo le'u" has the same meaning.

  However, rule 5 makes a special case for "zo le'u" :

    la'e lu zo le'u li'u -> the word "le'u"

    la'e lo'u zo le'u le'u -> ???


Point 11.

  I see that ZOI is processed before ZO:

  "zo zoi ti ta ti == zo (zoi ti ta ti)" ?


Point 12.

  SA is processed after LOhU ... LEhU, so I would assume that when we 
have :

      broda lo'u sa broda le'u

  nothing is erased.

  (Proper erasing with "broda lo'u le'u sa broda", right ?)


Now, my example rephrasing for the rules:

1a: .Y. (not followed by BU) -> whitespace
1b: (the formal rule is OK)
1c. After this rule, no .Y. remains for consideration except where 
there is Y+BU.

2a: ZOI (ignore Y) (whatever) ... (whatever) -> quote with 4 "words"
    ZOI (ignore Y) SI -> whitespace
    ZOI (ignore Y) SA ... -> whitespace
    ZOI (ignore Y) SU -> whitespace
2b: Unless immediately erased with SI, SA or SU, ZOI groups with the 
following word (delimiter), text and delimiter again. The entire group 
has 4 parts erasable by SI.
2c. After this rule, ZOI occurences have lost their value and have been 
grouped as quotes (together with the delimiters and content) and/or 
erased.


3a. ZO (anything) -> quote with 2 "words"
3b. (the formal rule is OK)
3c. After this point, ZO occurences have lost their value and have been 
grouped as quotes.

4a. (anything before) FAhO-not-in-quote -> end of text
4b. (the formal rule is OK)
4c. After this point, FAhO occurences have lost their value, and 
everything after the first has disappeared.

5a. LOhU (lojban words) LEhU -> quote with number of lojban words 
present in the middle + 2
5b. LOhU...LEhU quotes any string of valid lojban words, including LEhU 
itself if prefixed with ZO.
5c. After this point, LOhU...LEhU occurences and everything in between 
have lost their value and have been grouped as quotes.

Note about 5: no need to state a special case for ZO+LEhU, since rule 3 
has processed ZO+LEhU already.

6. (nothing to say)

7a. (whatever) (ignore Y) SI -> whitespace
7b. (the formal rule is OK)
7c. After this point, SI occurences and the word before them have been 
erased and lost their value.

Note about 7: no need to state a special case for ZO+SI, since rule 3 
has processed ZO+SI already.

8a. (whatever X) (lojban words) SA (ignore Y) (whatever same X) -> 
(whatever X)
8b. SA erases itself and anything that precedes up to a word of same 
selma'o as the word that follows.
8c. After this point, SA occurences and the text before them (back to 
and including the delimiter) have been erased and lost their value.

9a. (anything) SU -> whitespace
9b. SU erases itself and everything before it.
9c. After this point, SU occurences and the text before them have been 
erased and lost their value.

10a. (whatever) (ignore Y and BAhE) ZEI (whatever) -> lujvo
10b. ZEI groups the word just before (except BAhE) and just after into 
a lujvo.
10c. After this point, ZEI occurences and the thing before and after 
them have been grouped into lujvo and lost their value.

Note about 10: no need to state the special cases about ZOI, ZO, LOhU, 
LEhU, SI, SA and SU since those have been processed by previous rules.

11a. (whatever) (ignore BAhE) BU -> lerfu
11b. BU groups with the word just before (except BAhE) into a lerfu 
word.
11c. After this point, BU occurences and the word before them have been 
grouped and lost their value.

Note about 11: no need to state the special cases about ZOI ZO LOhU 
LEhU SI SA SU and ZEI since those have been processed by previous 
rules.


Thanks for any answers / comments.

--
Raphaël

Attachment: PGP.sig
Description: =?ISO-8859-1?Q?Ceci_est_une_signature_=E9lectronique_PGP?=