Le 04/10/2013 14:14, Tudor Girba a écrit :
Hi,
On Fri, Oct 4, 2013 at 1:41 PM, Goubier Thierry <[email protected]
<mailto:[email protected]>> wrote:
Hi Tudor,
I found your visualisation very interesting, and wondered about one
thing linked to your blog post and if I got it right.
The visualisation you're showing is able to show, at the single
class level, if a class is more or less regular (i.e. tidy == well
designed?) but you show that it will loose very shortly it's tidy
shape if the classes it joins to are added to the visualisation. And
I wondered if this was a choice on the graph layout algorithm you've
choosen, i.e. that maintaining the tidiness of the class could be
done with a different layout algorithm? Maybe one which has a
measure of tidiness of the class at the local level, and weight it
against the position and attraction of its links to the other
components in the application?
My argumentation is that if we look at a class isolated it will have a
certain shape, and because we are exposed to this view (the IDE is
always showing me only that) all the time we will optimize it such that
the shape is tidy. But if we look at the class in the context of its
interactions the shape will be less recognizable. The chosen layout is a
very simple one, and the goal is not to say whether the structure is
good or bad. This is neither good nor bad. It simply argues how steering
the architecture has to take these forces into account.
Hum, my point would be that a locally well designed class is a factor to
strive for; also that the overall architecture has it's impact on it as
well, but not so as to so easily erase the local good property when
looking at the visualisation, otherwise the latter may not end up so
usefull, that's all.
I do agree with your point, I just believe the layout could carry a bit
more insight in trying to both show the local property the code was
trying to achieve (and areas where it failed) and the "emergence" of the
architecture coupling on it. Your representation seemed to imply that
local good design is unimportant in the architectural view, and I'm sure
this is not what it is supposed to convey.
In the mean time, I'm slowly discovering we have amazing software
architecture visualisation tools :)
They really are amazing :). If you want to get more information about
it, you can join the Moose mailing list:
http://www.moosetechnology.org/about/contact
I'll do once I have a strong/practical incentive :) I know its there,
Stéphane showed a lot of interesting things in it when he came here, but
I didn't managed to organize things around that subject.
Would really like to setup a project on that if I get the chance.
What do you mean? What kind of a project?
Collaborative R or R&D project :) Only way I can justify working on
something like that...
A good fit would be visualisation of large software in my lab expertise
area : high performance and real time massively parallel software on
embedded systems.
Thierry
--
Thierry Goubier
CEA list
Laboratoire des Fondations des Systèmes Temps Réel Embarqués
91191 Gif sur Yvette Cedex
France
Phone/Fax: +33 (0) 1 69 08 32 92 / 83 95
