3 ms·
I used to develop next-gen high-end OO CASE tools, and am big into systems analysis/design/etc., visual thinking and communication, etc. A few random comments,
by neilv 6y ago
I used to develop next-gen high-end OO CASE tools, and am big into systems analysis/design/etc., visual thinking and communication, etc.
A few random comments, on latest impressions of UML, while trying to get something done last week, but not a rigorous look UML...
I happened to skim parts of the UML spec (796-page PDF) last Thursday/Friday, having not seen UML in a while.
I appreciate all the goodness from various methodologists that was incorporated into UML. And, especially as a tools developer, I also appreciate the semi-formal metamodels, which have a few key uses, most of them for tools developers.
Unfortunately, the particular UML features I needed last Thu/Fri (for a cross-org high-level process&architecture model) didn't seem to be supported sufficiently well (or at all) by any of the first half-dozen UML tools I tried. (I tried open source, and then lightweight Web SaaS. People generally do that, maybe more SaaS first, and want to get started rapidly. I'm not going to casually install a huge closed CASE platform, without checking the SaaSes and open source.)
By Friday afternoon, finding all these tools hindered more than they helped, I said "fiddlesticks this barrier situation", and decided to just use our GSuite Draw. Although, on the surface, it appears to be pretty much 1980s shared-whiteboard simple drawing CSCW (with a little 1990 Visio diagramming innovations), I was able to express enough of what I needed without too much effort, and I was even able to reason from the diagram a little as I was manipulating it.
To make GSuite Draw viable for my immediate purpose of starting to work through a model, I did have to use a metamodel and notation that was easier to draw without special tool support, but it worked for this particular model.
I'd still use a specialized UML tool for the most familiar Rumbaugh-ian static object modeling. And, if the tool supported them well, I'd also use for Harel state modeling, maybe event traces, maybe Jacobson use case modeling, and some other UML ones (e.g., Activity Diagrams, especially with the object flows, not just flowchart control flows).
I wouldn't generalize much last week's experience, to adoption by developers in general, since I'm a fringe person who knows specific metamodel features that exist, which I want to select for a particular need -- rather than being a new person dropped in front of a particular kind of UML diagram, and told to start expressing my system in whatever terms it gives me.
I haven't done a rigorous survey of current method&tool practice and offerings, but I'm starting to form a suspicion that adoption and execution challenges for UML-ish things today are very similar to what they were 20 years ago. The markets certainly seem to have changed in some ways, but even some of those ways (e.g., more widespread framework reuse), were already being talked about for modeling 20 years ago.