[Uml-devel] CASE tools and IDEs

[Andrew Sutton]

since we're talking about integrating umbrello and gideon, i have to throw in 
my $.02

first, IDE's are overly compiler-centric. developers have to deal with files 
and directories because that's what the build environment understandes. the 
files don't really correspond to anything useful from the development 
standpoint. all we care about is class and implementation - i don't give a 
shit what file its in.

second, IDE's are good if all you care about is the code. sure, the class view 
is useful - i guess, but it doesn't provide a very good abstraction for 
viewing the project. neither does the file view. its just lists of classes 
and files and the like.

to sum up, i think that IDE's will be obsolete - at least in their current 
form sometime in the not-so-distant future. from the other side of the 
argument, we have CASE tools - like UML modeler's and such.

traditionally, they haven't really been to useful for development because 
they've not been tied to the IDE. also, roundtrip engineering (making a 
change to the code affects the design in real time and vice versa) hasn't 
been implemented very well. additionally, CASE tools don't provide really 
good mechanisms for navigating the possible abstractions of your project. 
usually, there's a tree view with all your classes and use cases and 
diagrams, but that's pretty weak. UML allows for significantly more 
complicated abstractions - for example, viewing inheritance or containment 
hierarchies for a library.

what a lot of people don't know is that UML is designed to contain 
implementation as much as specification. in other words, the CASE tool could 
become the IDE. that, as much as anything is my real goal for Umbrello - to 
replace the standard IDE with a CASE tool. all that's missing is adaptations 
to UML to support the modeling of build environments, compiler options and 
the like. hmmm... that sounds like a master's thesis.

as for Umbrello and Gideon - i don't see it happening - certainly not for 1.x. 
if somebody wants to adapt the design for 2.0 to support roundtrip 
engineering, its all them - as long as it doesn't break the design for the 
rest of the application and doesn't steer umbrello away from what we want it 
to be.

andy