[freearchitecture] Re: x3d-cad vs. step

[Eric Wilhelm]

# from Lars Grobe
# on Tuesday 09 May 2006 12:40 am:

>1) A central "model server" holds the model, and apps connect to it
> via a defined interface. The connected apps are notified when changes
> are made and can update than. So the actual apps are like browsers ,
> as you described them, and it would just be necessary to add the
> interfaces. One idea would be the verse protocol, but I am not sure
> if it fits in such a centralized approach. 

This is actually one of the long-term goals of VectorSection.  Because 
entities are designed to be able to be atomic bits on the filesystem, a 
daemon could handle concurrency and network issues on top of that.

> The UI inconsistency is still not solved.

UI inconsistency is going to be difficult without a company like Apple 
building everything.  Everyone has their own ideas and demands for 
interaction design.  It doesn't help that most existing GUI toolkits 
both impose an interaction paradigm and offer lots of "convenience" 
functions that don't directly relate to the GUI.  See pythoncad's 
attempt to have alternative interfaces.  I won't say that that has 
failed, but the Mac interface has certainly stalled and it takes a lot 
of pseudo-repeated code to create two interfaces.  But, that's just to 
change mostly the appearance.  I don't know of an example where anyone 
has implemented a switch between e.g. an autocad-style command line and 
a discoverable heads-up interaction model in one program.  That 
paradigm-switching stuff is hard.

>2) We develop a new UI interface + model holding application and try
> to get "engines" derived from existing projects to plug-in. So we had
> ONE UI, and specialiced tools (invisible for the user).

Again, a big job.  Trying to create an API that allows any application 
to get its work done while limiting the UI is going to lead to huge 
amounts of metaphor shear.  IMO, that's not the place for an 
abstraction layer (see WxWidgets for a trivial (albeit inverse) example 
of why this is both difficult and limiting.)

I do like the idea of specialized tools fitting into a framework.  I'm 
hoping that VectorSection can lead to that without requiring that those 
tools be usable only from a GUI.

>3) We just promote a common file format and UI style among projects
> and see how well they will fit together. The Linux Standards Base is
> a bit like that. People will have to install a collection of tools
> and learn to use them.

Hmm.  Common format.  That's a great idea.

--Eric
-- 
Issues of control, repair, improvement, cost, or just plain
understandability all come down strongly in favor of open source
solutions to complex problems of any sort.
--Robert G. Brown
---------------------------------------------------
    http://scratchcomputing.com
---------------------------------------------------