[Opengc-devel] opengc "toolkit" proposal

[Damion S. <be...@cs...>]

Hi folks,

I've been doing some thinking over the past few days about how to 
address the future of OpenGC (both in terms of where it's going in 
functionality and development). I think I'll throw the following out as 
a straw man proposal to get feedback on. To summarize, since the 
following may be a bit long-winded: OpenGC will be rewritten into a 
form of which I am the sole author (or have express written permission 
to include the code of others), after which it will be released as open 
source under a license modeled after the VTK license.

1) I will completely rewrite OpenGC based in part on my source code 
from pre Dec. 2001 (the date of the first CVS contributions of others), 
in part on modern source code about which there is no dispute regarding 
sole authorship (the font manager, X-Plane data source, the NEW nav 
database, etc.), and in part on a new design I've been working on based 
on the ITK and VTK toolkits, http://www.itk.org and http://www.vtk.org 
respectively.

2) Under this new design, the appobject would disappear completely, as 
would the gauge and gaugecomponent classes (to be replaced by a more 
general widget class). Objects would also be reference counted, using 
smart pointers, to avoid memory leaks. The eventual goal would be to 
provide something along the lines of VTK; i.e., a finished 
"application" as such is not provided, but the tools to rapidly produce 
one are. Example applications would of course be included, but more 
full-featured applications (with advanced mouse interaction for 
instance) would be left as an exercise for the reader. Under this 
model, an application main() might look something like:

main()
{
	ogcRenderer* pRen = ogcRenderer::New();
	ogcDataSource* pDataSource = ogcFSUIPCDataSource::New();

	ogc777PFD* pPFD = ogc777PFD::New();
	pPFD->SetDataSource(pDataSource);
	pRen->AddWidget(pPFD);
}

The above code may look a bit odd to those of you not familiar with 
VTK, but is intended to show the basic steps involved in creating a 
simple one-gauge display that drives a 777 PFD from FSUIPC. The _major_ 
advantages of switching OpenGC to this modular approach are:

A) It becomes possible to use the CABLE tool ( 
http://www.itk.org/HTML/CableSwig.html ) to generate wrappings in other 
languages. Prefer to code your GUI in Java or Tcl? No problem. This 
would address the common question of "can I use OpenGC in language X".

B) OpenGC is not tied into a particular application framework. As it 
stands, OpenGC doesn't really exist without AppObject and main.cpp - it 
would be nice if this weren't true.

3) This code would be distributed under a modified form of the Kitware 
license/VTK toolkit copyright. This is an extremely open license (do 
whatever you want), with one exception: the version of _contributed_ 
code placed in the repository becomes copyrighted solely by myself. All 
users and developers still retain full rights to use and redistribute 
the code, with or without modifications, with or without included 
source.

I've talked to Kitware in the past, and they have adopted this license 
purely to avoid copyright dispute regarding the code INSIDE the 
repository; as the license states, users are still free to do WHATEVER 
THEY WANT with the code. Please take a look at the license here: 
http://www.vtk.org/copyright.php

This is likely to be the portion of this proposal that creates the most 
angst among others, so feedback is encouraged. Please keep in mind that 
this license has worked extremely well for Kitware; VTK has something 
on the order of 2000 active developers. The copyright clause does not 
mean that you're giving up copyright to code you've written; you're 
free to publish another version elsewhere copyrighted solely by 
yourself.

4) As I produce a "sanitized" version of OpenGC, it will posted for 
everyone to see. I do not want to be accused of using code which 
someone else may legitimately have a claim to, so the entire revision 
process will be conducted in the open. After it has been established 
that I am the sole author of version "X" of OpenGC, it will again be 
opened up for public contributions under the new license terms.

5) Code for which permission cannot be obtained for inclusion in OpenGC 
mark II will have to be removed from the repository. I will try and 
work with the FlightGear folks to produce a non-GPL'd version of the 
FlightGear data source. All of the data sources will be changing 
dramatically, both in function and variable names. The current version 
is very poorly designed, by my own admission, and does not allow any 
form of bi-directional communication.

Given how well things have worked out for the Kitware folks and the 
users of VTK (of which I am one) I think this is probably the best 
direction in which to proceed. I realize that this will effectively 
halt new development of OpenGC for 4-6 weeks during the rewrite, but 
I'd rather get this sorted out sooner rather than later. The current 
version of OpenGC, as licensed under the GPL, will continue to be 
available for download even after the new version is released.

Feedback is encouraged...

Cheers,
-Damion-

---------
Damion Shelton
The Open Source Glass Cockpit Project (OpenGC)
Carnegie Mellon University, Robotics Institute
http://www.opengc.org
da...@op...