Le 22 avr. 05, à 18:07, Alex Perez a écrit :
And, if I understand correctly, in the ideal concept of how Etoile
would work, there are only services, no applications. How does that
impact what you're describing?
Not exactly. There will be applications, because you must have them.
But they will be VERY light, and almost everything will be implemented
in frameworks/reusable libraries, not services (Although there will be
plenty of those as well, as far as I understand it). Anyways, from
what I read of your (otherwise sound) idea, there would be no reason
to not simply have "ShareWith" be implemented as a service.
I must reiterate my opinion: I don't think we should *only* have
components. Not just because of the technical reasons (I believe that's
what implies Alex by "you must have them" in talking about apps) but
because if we just provides components and leave the user the charge of
combining the components for everything, I don't think it will work as
an environment: it will be too complex for most users, even if
powerful, and that's not what we want.
Basically, combining components ends up creating "applications" anyway.
And sometimes you don't work with documents, and thus the "filter"
mechanism or composition mechanism often touted for components just
doesn't work -- sometimes, you really use an application that doesn't
create things, but that do things..
So in my view, we will still have applications, if only as ready-to-go
packages of components. Example: think of a garageband-like
application: you can imagine having lots of components to register
sound, play it, modify it, play midi, etc. And an user could then use
theses components to do what he will do with garageband. But the UI
will truely, absolutely, sucks. So we can't just say "yadda yadda, here
is thousands of components, use them, mix them, do what you want". In
my opinion, what *should* be done is to provides a "skeleton"
application that will provides them with an integrated UI (say, similar
to garageband). This application will of course use the available
system components, and will be easily modifiable and extendable. But to
the user it will be an application. And anyway he will be able to use
the components directly, if he wants; but well..
So effectively I think we won't completely abandon "applications" --
but thoses applications will indeed be very light and will extensively
use components for as much uses as we can think of. And of course, when
it's logicial, the environment should be focused on documents.
The "pure" components approach would still be possible (as in, here is
a blank canvas, here is components, here is a way to connect them, now
play with them), and it will be useful too. I just don't think it's
enough.
--
Nicolas Roard
"Any sufficiently advanced technology is indistinguishable from magic."
-Arthur C. Clarke
