Philippe wrote:

<<If you use "purely graphical languages", that you
don't use "phonetics", but writing, or drawing.
"Phone" (phi, omega, nu, eta) means "voice" in Greek.>>

It's been firmly established for probably thirty years now that "phonetics"
and "phonology" don't have to apply
to sounds.   Why would anyone say this?   Simple: Sign Language.   There are
lots of papers on the phonology of
sign language.   I attended several talks, in fact, on the phonetics and
phonology of ASL just last quarter (our
department was doing a job search for a sign language research position).
So what this means is that we know
longer have to depend on the definition of a Greek term to know the
definition of phonetics.   Phonetics is the
study of the implements (natural or unnatural) used to communicate.   In sign
language, the active articulators
are the fingers, hands, arms and face (though there's also an argument that
the torso is used).   Using this
definition, the active articulators of a purely graphic language would be the
things used to write the language
down.   If this is a computer, the study becomes less interesting.   However,
if you were using plain old pencil
and paper to write, then the phonetics of a written language would be the
changes in the glyphs.   So, for
example, if the word for "circle" was a drawn circle, but the word for "oval"
was an oval, it would be interesting
to see just how ovular you could get before an interpretation of "circle"
would become impossible--in other
words, what the boundaries for simple phonetic variation are.   This would be
no different from saying that, to
pronounce "pipe" you could say [p_hajp] or [p_ha:p] or [pajp], or maybe even
[pap], but definitely not [faip] or
something totally off the way like [flut].   How one draws a glyph is
*exactly* the same phenomenon.

Continuing (snipping some):
<<If you do [writing, phonetics and word definition] first, you're
almost certain to have to do the whole thing again,
because it won't work.>>

Based purely on your analogy, I'd say, "Wow, yeah!   You've made a convincing
analogy.   Therefore, what you
say must be true."   However, I know it to be false.   What's one to do?

I always start out with the phonology, phonetics, and writing system of a
language.   I then use this to build the
rest of the language.   In fact, I think up phonological phenomena I find
interesting, and then set up a syntax or
morphology that can showcase the phenomena.   Doing it the other way around
would be incredibly dull.   But
maybe this is the difference between the computer scientist/programmer and
the English major (me).   I don't
see the distinction as being important: I simply prefer one way over the
other.   You, however, have made the
claim that if you do it my way, you're basically fated to do it your way,
where the first round of phonology, etc.
was just practice, and could in no way work with the rest of the language.
A bold claim.   I think there are
probably a lot of phonology, etc.-first conlangers on the list, and I bet
that not all of them were forced to do a
total teardown of the phonology once they created a syntax.

Just a note: this, of course, excludes the possibility of doing it all at the
same time.   That is, getting an idea for a
particular syntactic/morphological phenomenon, setting up a basic phonology
to test it, refining the specific
phenomenon, then refining the phonology, then moving on to everything else.
Though systematicity has its
perks, I'd wager that there are a lot of conlangers who do it this way, as
well, in which case a chicken or egg
argument would be largely irrelevant.

"sunly eleSkarez ygralleryf ydZZixelje je ox2mejze."
"No eternal reward will forgive us now for wasting the dawn."

-Jim Morrison