Re: UML cooperation

[vmilitaru]

* Alejandro Imass <address@hidden> [001026 08:16]:
> Well. We must admit that DIA is VERY limited at the
> moment and it's a pain to use. BUT (and it's a BIG AND
> LOUD BUT) the framework on which DIA is built will
> assure not only continuity (GIMP LIB, GNOME, XML,
> etc..) but I can bet anyone the it will be far more
> functional than any other OO package (like RR) in the
> "near" future. For this reason I placed my bets on it
> and am using it despite it's present (BTW I'm only in
> 0.8 and probably many things are fixed in 0.86) state.

I wouldn't want this to turn into a religious argument. When the new "Dia on 
steroids" will outfeature and outperform RR, I'll be the first to dump RR 
(although I presume RR will not stand idle, but will add other convincing 
arguments to justify its pricetag, in the time Dia will take to catch up). I am 
concerned presently with "here" and "now": for the current task, what are the 
best tools at my disposal Today ! 

Remember the computer world "before" the arrival of the GUI ?
What I see from RR (perhaps other tools I am not aware of) is the emergence of 
a new development paradigm, where the focus shifts from editing source code to 
a higher graphical level (the UML diagrams) WHILE maintaining performance (!). 
Yes, there are other higher level tools, like the scripting languages, which 
are also very inefficient (read "slow"). What RR allows is higher level (and 
graphical) development in the form of UML diagrams !!AND!! no loss in 
performance (in its ability to generate source code for efficient compiled 
languages). 

So, this new dev paradigm standardizes on UML, with "legacy" compiled languages 
(C++, etc) as an interim state. RR's appeal is not in its drawing abilities, 
but in introducing a new development cycle: Legacy-source-code -> UML-diagrams 
-> many-lang-source-code -> binary-executable. Dia should not just replicate 
RR's revolutionary approach (at least to me it looks revolutionary), but 
integrate it in a seamless development flow, where developers' resources will 
be focussed at the UML level, not at the source-code level. 

Granted, RR present incantation's ability are far from a flawless 
implementation of this vision (it only generates skeletal source code, for one, 
no executable), but it is gradually improving, definitely moving in that 
direction. Perhaps this wouldn't be such a farfetched idea for the open sorce 
community to also start looking in this direction, maybe put some resources in 
this potential upcoming battle (from my point of view). And possibly a major 
one, if the implications sketched in this message materialize (for example, 
expect Microsoft to buy Rational,Inc and integrate Rose into VisualStudio, if 
the dev paradigm described here starts to get serious momentum ... or if they 
read this e-mail - ha, ha; like with the internet, MS won't see it coming, but 
once it focusses its formidable resources on something, watch out for bodies 
falling; just read that they bought a third of Corel, a war trophy, no doubt).

But coming back to Dia vs. RR, it's more than the drawing ...

So I guess that's my opinion on the subject.
Vio