Thanks for the clarifications, Stefan.
They help, but the main strategic worry is still hanging:
Open modelling in Julia must not become collateral damage in the push for a slick, vendor-controlled ecosystem.
Wouldn’t right now every fresh component library have to run through Dyad if we want
- the shiny SVG GUI
- the “official” standard libraries
- first-class Julia/SciML hooks?
In Modelica I have a choice of proprietary and open tools; competition can “do its thing” in principle.
To me the lock-in worries still remain valid:
- Spec—Without a public spec (or at least a living “grammar” documentation) how can anyone ever write an open compiler?
- Open ModelingToolkit libraries—What stops today’s MIT/BSD ModelingToolkit libraries from quietly sliding behind that paywall?
- Governance—How do we make sure Dyad complements, but not dictates the future of Julia modelling?
- FOSS author seat—Is there any chance for seat fees getting waived (or heavily discounted) for developers who ship their libraries under OSI licences?
See also:
As a reminder what to watch out for (edit):
Back in 2022, Cory Doctorow coined the term “enshittification” to describe a cycle that has played out again and again in the online economy. Entrepreneurs start off making high-minded promises to get new users to try their platforms. But once users, vendors, and advertisers have been locked in—by network effects, insurmountable collective action problems, high switching costs—the tactics change. The platform owners start squeezing their users for everything they can get, even as the platform fills with ever more low-quality slop. Then they start squeezing vendors and advertisers too.
Source. And probably everybody promises not to do it beforehand?