[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Moonlight|3D (was: Re: artists)

Am Samstag, 30. August 2003 04:45 schrieben Sie:
> Gregor Mückl wrote:
> > The alternative is to help people out with writing an alternative to
> > blender. I've assembled a team who could do it. We're preparing to start
> > coding now and might have a first usable version within a year.
> Writing a complete 3D modeller is an incredibly hard problem.  I'd be quite
> suprised if you could get anything out the door within a year from a
> standing start.
> I worked with the PrettyPoly group (now defunct).  We worked long and hard
> - but the enthusiasm ran out before the work did.
> My advice:
>    1) Implement 'undo' from day one - don't put it off until later.

I believe that we got something even better than any undo: The user is 
creating a tree of operators which in turn create the final scene. This tree 
is editable at any time during that process. Nodes can be added and removed 
at any time. This is much more powerful than a simple undo/redo mechanism and 
is in fact the heart of the overall design.

I noticed quite early that the undo/redo features are something that you can 
use as a criterium to estimate the real power of the tool. Big packages seem 
to have implemented this by design while small ones miss it or have some 
simple "keep the previous states in a stack" approach layered all over the 

>    2) Take a tip from GIMP and do everything with scripting.  Build a core
>       editing engine, a core rendering engine and a GUI - but connect the
> three together with scripting - you won't regret it.

Hm... scripting would be nice. I think we can hook script interpreters into 
the system at two points. The system is desigend to be modular and 
distributable across a network. So a single script cannot handle both the GUI 
and the actual editing operation. But anyway, I'll keep that in mind.

>    3) Make sure everyone knows how much work it is - don't kid yourselves.
>       It's going to take two years to get a usable modeller out the door.

One of our team has written a simple editor for his thesis (which was about 
particle systems). He knows and he keeps telling us. And I think that's why 
he wants to focus on the GUI :-). Anyway, we have something worse before 
that: The design demands a basic framework which in itself will need at the 
very least about 5.000 lines of code. Without it there's no GUI, no meshrep, 
no nothing.

The really hard part seems to be to get it to a state in which users can start 
to adopt it and start to complain about missing features. I claim that this 
is the breakthrough point for any open source project.

Anyway, thank you all for your interest. I hope we can meet your expectations.