Hi Erwin,
System in Cloud Modeler is the entry part of the solution I am creating. Inspiration for the initiative was my work on video processing algorithms. I've seen a gap between idea of domain engineers and the implementation to working, market-ready application. My business model is to offer fully-feature Modeler for free. Second part of my solution is cloud service via which models can be executed on infrastructure of cloud providers of choice.
In long term I want to open my Modeler. I understand that if you want to something be a standard, it has to be open-source. But this step I decided to pass after cloud service reach general availability.
I have no ambition to make System in Cloud alone standard. My Modeler creates abstraction for any kind of computations. One of the main principles of Modeler's engine is to keep abstraction (workflow organization) overhead as small as possible. Probably it would be not adaptable for those who look for simulation tool.
To your questions:
+ Did you implement an own "process engine" or are you using one of the existing open-source ones?
I have developed own process engine. It is because I had a clear vision of what rules I wanted to have. If I could, I preferred to create a tool that solves my problems than adapt my problems to tool. E.g.:
* Task can be executed if a group of ports have data.
* Task can be a generator that is periodically activated.
* Task can ask to be activated by its own if for example it received data from external world.
* Data in ports can be in queue.
* Data in ports can be lost if Task is not processing fast enough or previous Task in chain can be frozen for a while.
* and some others.
+ I would be interested to know on what underlying technologies you've built your modeler.
At the beginning of my development I was inspired by Activiti project (http://www.activiti.org/) and its Designer. It is why I chose: Graphiti. For Models EMF. And for Modeler it would be almost all I think. Engine is written from scratch in Java 7.
+ So System in Cloud could provide an alternative flow engine to orchestrate Tasks?
Today engine is interpreting one type of defined model stored in xmi. I don't say that in the future I would not introduce the possibility of other storage. Some human readable could be useful.
+ supporting services and APIs. Maybe you want to collaborate on the requirements and implementations in this space?
I think I don't understand well what is about. But if it is about something outside non-core modules maybe we could develop them together for both projects.
Contribution:
+ Depending on your wish/interest, you could contribute and/or integrate (parts of) System in Cloud in the Triquetrum project?
+ Or a more light-weight approach could be to share work between the two projects?
With source I would wait up to time when Modeler is open. Moreover, I created proper engine and diagrams have their own semantics (e.g. sync and async ports). Probably it would require a lot of work to make two solution compatible.
What is sure that I can offer: my experience and knowledge gained during development. I will start to be active in Graphiti forum once again. If you would like to know how I made certain Graphiti things you can post a message there and ping me about it. You can also write to me directly if you think that the question and response would not be generic.
I also can share information about non-Graphiti things like what technologies has been used and how I solved some problems (e.g. python integration).
I am open to redefine my role in science group in the future. I really believe in open science and I very appreciate what you are doing in Science WG.
Marek
