critique of gua\spi
NB. In the text below, "I" refers to John Cowan, and "you" refers to Jim Carter. --And Rosta
The following is an unorganized series of notes commenting on -gua!spi
characteristics I have problems with. Much of this, of course, reflects
Lojban biases. (It's about time somebody ran your stunt the other way zo'o.)
Character set: I find the use of c, j, q, x rebarbative. In particular,
x = sh in some languages, notably Portuguese And Pinyin, plus it's ch in catalan - .kreig.daniyl., and I doubt it has the
-gua!spi value in any language. Except some dialects of Basque baeba. Likewise q = ch in Pinyin.
Phonology: Affricates conflict with clusters; how can you reliably distinguish
between c and the cluster tq (resp. j and dx)?
Phonotactics (not explicitly discussed in the report): I find the consonant
clusters way too loose: many of them are unreasonably difficult to pronounce.
I would suggest starting from the Lojban set
and adding carefully and selectively. What is the existing phonotactics
Tones: Six is too damn many. Tell that to the Yue (cantonese) speakers. Four (high, low, rising, falling) would be
more like it. This would require a little more verbosity in subordinate
clauses and transitive compounds. In addition, it would be useful if
every monosyllabic word had an alternative disyllabic version to make
rising and falling tones easier, i.e. no "VV" vs "V" contrasts where the
V's are identical. (This may already be true, but should be guaranteed.)
Extensionality: Defining predicates purely in terms of the result sets
has known problems. In particular, "x1 has a heart" and "x1 has kidneys"
are true for the selfsame set of x1s, so in -gua!spi terms they are the
same predicate. But there would be no need to distinguish them anyway if there is nothing known that fits one predicate but not the other. For mathematical predicates, this may be all right (to
be negative is the same as to be less than zero) but in the real world
it can lead to annoying results. See Quine.
Macro-like pronouns: These barf on self-referential sentences and also on the
- The boy who deserves it will get the prize he wants.
Vanilla macroexpansion produces:
- The boy who deserves (the prize he wants) will get the prize (the boy who deserves it) wants.
- The boy who deserves (the prize (the boy who deserves it) wants) will get the prize (the boy who deserves (the prize he wants] wants.
And so on. You lose.
Compounds: You don't give any examples in the report of compounds of length
greater than two. Transitive compounds and event compounds can both suffer
from ambiguity about associativity.
Masses: As you know, I have problems with your identification of masses
with sets. The set of rats is large if there are many rats, whereas (part of)
the mass of rats is large if at least one rat is large. Masses may have
properties that appear self-contradictory because they are the conjuncts of
the properties of their components (the mass of you is green if at least one
of your components is green). Sets, on the other hand, have properties totally independent
of the properties of their members.
Negation: I believe you need something equivalent to
Lojban's na'e -- nonspecific scalar negation. lo na'e gerku is a non-dog,
probably another kind of animal (or something else if a different scale is
presumed). This is distinct from lo na gerku -- that which is not a dog,
contradiction pure and simple. There also seems to be no way to express
a metalinguistic negation (presupposition failure, categorical denial, etc.)
like Lojban na'i.
Parenthesis: It would be helpful to be able to have subordinate clauses
which are marked as digressions -- () in English prose -- or editorial
Erase all: A sentence start word which erased all of the speaker's previous
sentences (Lojban su) would be useful.
Change of topic: Sentence start words to clear anaphora assignments (Lojban
da'o) and sticky tenses (approximately Lojban ki and/or fu'o) would also
Subscripting: There should be a mechanism for attaching an arbitrary phrase
(typically a number) as a subscript to any word without affecting the word's
grammar otherwise. This is excellent for manufacturing extra variables and
Emphasis: Lojban ba'e emphasizes the next word, bringing it to the speaker's
particular attention (similar to moving it to the beginning or the end of
its bridi). It is like italics in English prose.
Sentence anaphora: There seems to be no analogue of Lojban di'u and friends:
"the previous (following, earlier, later) utterance considered as an object".
This is distinct from go'i and friends, which repeat the utterance implicitly.
And after making all these changes you end up with Lojban with tones. And rafsi pretending to be gismu.