Message-ID: <421E846C.6070905@lojban.org>
Date: Thu, 24 Feb 2005 20:50:36 -0500
User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913)
MIME-Version: 1.0
To: Lojban List <lojban@yahoogroups.com>
References: <200502130829.09923.phma@phma.hn.org> <420F961A.40000@lojban.org>
In-Reply-To: <420F961A.40000@lojban.org>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
From: Bob LeChevalier <lojbab@lojban.org>
Subject: Re: [lojban] Archivist: Counterexamples to general fu'ivla lujvo
 making
Content-Length: 7793
Lines: 178

Bob LeChevalier wrote:
> Pierre Abbat wrote:
> 
>> The Book says: "Many attempts were made to add general mechanisms for 
>> making lujvo that contained fu'ivla, but all failed on obvious or 
>> obscure counterexamples; finally the general ``zei'' mechanism was 
>> devised instead." Xorxes has come up with a morphology that purports 
>> to make lujvo from all fu'ivla and even cmene. Can someone dig up the 
>> counterexamples?
>>  
>>
> I  doubt if there is any record of counterexamples unless they are in 
> the Lojban List archives.  Nora was the one who kept finding 
> counterexamples, but I was posting for her in those days.  I vaguely 
> associate the proposed introduction of "zei" with a LogFest discussion 
> between Nora, John Cowan and perhaps others; of course there were no 
> meaningful permanent records kept of Logfest discussions.
> 
> If it helps in searching archives, I believe that the predecessor to zei 
> was a linking "hyphen", "iy".  I can't remember exactly how it was glued 
> onto the components.
> 
> My archived notes on the morphology algorithm show that "iy" was added 
> sometime between July and Sept 1991, probably at that year's LogFest.  I 
> think it survived a year or two before further work showed it was broken 
> and ZEI was added.  So to look for counterexamples, I would search list 
> archives between July 1991 and perhaps Sept 1994 looking for keywords 
> "morphology", "le'avla" and "fu'ivla", and text string "iy" which should 
> not be too common in Lojban List except in such a discussion.
> 
> Sorry if that isn't sufficient.

I just found something in my notes that addresses the replacement of iy 
by zei, showing the reasoning, part of something else that may or may 
not have been posted to Lojban List about the date that was indicated in 
the text, where someone might be able to find more.  It doesn't exactly 
have "counterexamples" but it does clearly have the rationale why we 
gave up on iy  (some of the other things discussed could be relevant as 
well)  See under proposal 2 below

>    >Subject: Morphology proposal #3:  tosrmabru, or hyphen-buffer equivalence
> 
> Nora understands this one to be withdrawn.  It was thus not presented to
> le ta'agri for comment.
> 
> My suspicion is that the buffer, and maybe the hyphen as well, will have
> a length difference as well as vocalic character.  Until we get a few
> people who can talk at fluent speeds, the problems are purely
> theoretical.  And's point is taken, and we know what to watch for.
> 
>    Subject: Morphology proposal 2:  super-glue
>    Date: Mon, 18 Nov 91 14:29:59 EST
> 
>    CURRENT LANGUAGE:
> 
>    Currently "iy" is used to make le'avla lujvo by a set of rules that
>    involves placing "iy" both before and after a lea'vla (unless it is
>    initial, final, or adjoins another lea'vla) and then merging it with
>    arbitrary other rafsi or lujvo, with some exceptions which are implicit
>    in Nora's algorithm but haven't been nailed down anywhere.
> 
> Not yet adopted into the current language. le'avla and their
> idiosyncrasies were omitted from the o/p/m baseline as not yet defined.
> 
>    PROPOSED CHANGE:
> 
>    Use "iy" as superglue for combining any list of words (none of which
>    contain "iy") into a single word.
> 
> le ta'agri consensus recommends assigning a normal cmavo (not iy) for this
> purpose.  The result will not be a single word morphologically, but will
> have the semantics of a lujvo.  By this we mean that stress rules applicable
> apply to each component independently, and "si" erases a single component
> word.
> 
> The new cmavo should preferably be a monosyllable (I don't have the list
> of free ones handy, but I think there is one).
> 
> The exact hierarchy of this rule in the metagrammar wrt quotes, erasers,
> UI, BAhE, must be specified in the grammar algorithm.
> 
> This is therefore to be considered a grammar change rather than a morphology
> change, and dealt with accordingly.
> 
> Note that regular lujvo can be described as quasi-lujvo as well, giving still
> one more form for a lujvo.
> 
> This collection of modifications seemed to make the concepts and rules most
> teachable to the learning non-expert Lojbanist who would want such rules.
> 
>    No rafsi are allowed, since rafsi are not words.  "citiyplatipusa" would
>    become "citno,iy,platipusa".
> 
> By not using "iy", we retain the option of restoring le'avla lujvo under
> some form of shortening algorithm, for use with Type 4 le'avla, and
> generally made only be people who are familiar with whatever complexity
> of rules is applicable.  Until we have a significant number of such
> le'avla, the Zipfean need for shortening the lujvo is of secondary
> import.  By then also, we may have better means for testing changes to
> the hopefully approved morphology algorithm.
> 
>    RATIONALE:
> 
>    1) The formation rules are very simple:  merge any set of words together.
> 
>    2) The analysis is also simple, and even allows secondary stresses on the
>    components, as /cItno,iy,platipUsa/ will be unambiguous.  If a stress is
>    detected, look two syllables ahead for a "iy" glue, and if found ignore
>    this stress.
> 
> This would be hard to explain clearly if you call the whole mess a
> single word.  You end up having words and WORDS.  Better to just have words.
> 
>    3) Many cmavo that do not have rafsi might be useful in compounds nonetheless,
>    with such possibilities as "X-ray", "gamma-ray", and "non-A, non-B hepatitis".
>    Since "iy" cannot begin a word, a single cmavo cannot fall off the front, so
>    "xy,iy,kantu", "ge'o,iy,gy,iy,kantu", and "na'e,iy,a,iy,bu,iy,je,iy,na'e,iy,
>    by,iy,xepatitisi" are valid brivla.
> 
>    In effect, this creates a new category of brivla, rather than expanding the
>    category of lujvo:  we now have gismu, lujvo, le'avla, and perhaps
>    "mi'avla" (combined-words).  Like "le'avla", this lujvo omits a "sel-".
> 
> Would "bridi,iy,valsi" mean the same as "brivla"?  If so, then they are
> another form of lujvo.  I (lojbab, not in consultation) would thus
> suggest and favor a compound of lujvo (se'ijvo?).
> 
>    Subject: Morphology proposal 1:  oi-oi-oi!
>    Date: Mon, 18 Nov 91 14:15:41 EST
> 
>    CURRENT LANGUAGE:
> 
>    Every cmavo that begins with a vowel must be preceded by pause.
> 
>    PROPOSED CHANGE:
> 
>    Allow strings of VV cmavo without a separating pause.  Nora's version of
>    the morphology is amended to include the sentence:
> 
>    "If a piece consists only of vowels and possible apostrophes, divide it left
>    to right into VV or V'V cmavo."
> 
> Nora's opinion, concurred in by piro lei ta'agri, was that there were a
> lot of loose ends to be tied up in this.
> 
> I interpolate from comments that eliminating ambiguous voiced glides
> completely in any fully controlled part of the morphology (thus,
> interior of names possibly excluded), is desireable.  There is also the
> question of interaction between diphthongs and single vowels:
> 
> ?au,a   ?a,ua  ?ai,a  ?a,ia
> 
> Then there are boundaries between cmavo and longer words
> 
> ?laiatlstan  ?= lai .atlstan  or even ?lai .iatlstan
> 
> on the other end  ?XAgraia ?= xagrai .a  or ?= xagrai .ia  (not "xagra .ia")
> 
> but yet no problem with ?XAgreui ?= xagre ui
> 
> This gives some idea of the scope to be considered.  Please address in
> detail with examples in a revised proposal
> 
>    RATIONALE:
> 
>    General reasonableness, plus compatibility with NB3 Loglan.  Currently words
>    like ".oi,oi,oi" look like UNK cmavo, but it is most unlikely that any human
>    being would interpret this string as anything but ".oi.oi.oi".
> 
> Any problems with strings of vowels on the end of le'avla?           


lojbab