[Uml-devel] point of (un)interest

[Andrew Sutton]

On Wednesday 26 March 2003 6:41 pm, Andrew Sutton wrote:
> if any body's curious, i'm posting my software viz paper to my web site as
> its being drafted. it's tentatively titled "techniques for visualizing UML:
> a survey". i've only written 3 relatively incomplete pages, but some of the
> background is there...

if anybody else is curios, i just put up a newer and longer revision of the 
paper. now its called "visualization techniques for software modeling 
applications that employ uml: a survey". its still a working title :)

there's a big section on architectural framework views. an architectural 
framework view is basically some visualization that shows the organization of 
architectural views and their relationships. for example, umbrello has a use 
case view and a logical view. rose has a deployment view. these are all views 
of the same architectural description. we use a framework view to show how 
they are related. it turns out that these views are all part of something 
called the 4+1 view - which is actually a framework - or at least that's what 
i call them.

so, what bearing does this have on u2... well, it has several implications.
	1. that we explicitly support a view that shows the 4+1 or n+1 framework 
views as a navigational interface. it's good for browsing architecture 
descriptions.
	2. by supporting this, we need to support a system for dynamic architectural 
view construction and management. there's some basic requirements for this in 
the visualization section of the 4+1 view.
	3. that we could support multiple framework views. this implies that we have 
some mapping between different views in different frameworks to the same 
actual underlying parts of the model. shouldn't be too hard.

my personal preference would be 3, but that might be a lot of work - besides, 
how many people actually use other framework views. if rational gets away 
just supporting the 4+1 (actually the 2+1 in their case) and forcing that on 
their users, then a system that allows n+1 views should be just as good.

anyway... just thought i'd throw that out.

andy