[Bf-committers] > Undo system design

[Goran Kocov]

It's official :) After careful consideration I
definitely think that implementing a non-linear undo
system is not a good idea, since it would do more
damage then good to Blender's workflow. 

Yesterday I was oriented mostly towards the technical
aspects of the system, when I should have been
thinking parallely about the implications on the
workflow. Since one of the best aspects of Blender is
it's clean, fast, efficient and transparent workflow
(something that not many apps can say about
themselves), complicating the workflow with a feature
whose sole purpose is making the workflow better would
be paradoxal and potentially disasterous. I've always
liked the KISS (keep it simple, stupid) philosophy and
I like Blender even more for functioning by it. 

My mistake was trying to gain some of the flexibility
of procedural systems (Houdini's OPs) and
quasi-procedural systems (3DS Max's Modifiers) by
(ab)using the undo system. Even in applications with
good non-linear data flow design and visual
representation of the data flow (for ex. Houdini) the
workflow suffers somewhat and attempting to do this
with manipulating the undo system IMO would prove to
be very confusing for the user. 

I still think that non-linear undo systems have their
place, just not in an environment where the datasets
are closely interconnected. Parts of Blender where a
separate undo stack could still useful would be: the
text editor (one undo stack per script), the scenes
(one global undo stack per scene) and maybe the
sequencer with it's animation data (since I think the
sequencer is separate from everything else in the
project, both from technical and workflow
point-of-view). 

On the other hand, saving the undo stack with the
project would still be a very nice feature to have. 

Goran


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com