The long slow death of UML

wolandscat.net

4 pointsb-man4 comments

The UML standard is vastly over-complicated

In other words, it becomes a problem instead of a solution.

In theory you could build UML 2 and the extensions to UML 2 pretty cleanly based on the Meta Object Facility except when you look at MOF it has some UML 1 concepts in it which are not quite the same so the resulting system would have both the UML 1 and UML 2 concepts of "Class" in it.

If you dealt with all that you could possibly bootstrap an implementation of UML 2 on a simple core, but I think the UML vendors who backed the specification went out of their way to make this difficult because the last thing they'd want is for there to be a simple implementation of it.

MOF is only "simple" because it is lower level than UML proper. A lot of "not quite the same" would come to light if two different people attempted to rebuild the respective UML subsets they care about.

I'm sure of it. I worked on a project to turn a MOF-based representation of the ISO 20022 standard to RDF (as well as being able to write the messages in RDF) and I wrote some tooling to do it, looking at the specs it seemed it should have been possible to bootstrap a UML 2 system using the same approach but I saw enough strange discontinuities that I didn't go forward, but had I invested two months I probably would have solved the "two Class" problem and made some major progres. But boy have I done some many 2 month projects where I got to the bottom of something like that and bootstrapped a system and then found the world just didn't care... I'd have the most productive 2 months of my career and make $0.00.

Reply on news.ycombinator.com