The Model Driven Software Network
Raise your level of abstraction
Bottom-up Domain Specific Modeling
Hello all,
I want to announce my own approach to domain-specific model driven development. I call it 'bottom-up DSM (BU-DSM)' (because of modeling "motivation direction"). In reality it based on ideas that are very similar to Domain Specific Multimodeling Paradigm described by Anders Hessellund.
It's main ideas:
- using IoC-container XML configuration as 'final' model (so, model interpretation is used instead of code-generation). This enables components composition and interaction from different DSM and using these components even outside of DSM space (from custom code, for instance)
- using XSL-based rules for describing transformations from XML-based DSM models to 'final' model
- designing DSM specifications in respect to existent DSM-s, transformations, components and infrastructure. This means that modeling should start from 'bottom' (low-level components, low level DSM-s) and level of 'abstraction' (in domain direction) should be decreased iteratively (so usually DSM will be transformed into more abstract DSM in this way: M1 -> M2 -> M3 -> ... -> M (ioc XML config - abstract and universal model representation).
This approach was successfully used for building such systems like comindwork.com and sqaservices.com extranet since early 2007. It still documented not very well but I'm going to describe it as part of my dissertation work and some aspects already covered by articles.
As part of dissertation work implementation I've started open-source project (NReco) that supports ideas of BU-DSM and makes them real (of course it contains only simple examples - for number of reasons I cannot share real-life infrastructure used for commercial system development).
Feel free to ask you questions, I'll be glad to read your feedbacks.
I want to announce my own approach to domain-specific model driven development. I call it 'bottom-up DSM (BU-DSM)' (because of modeling "motivation direction"). In reality it based on ideas that are very similar to Domain Specific Multimodeling Paradigm described by Anders Hessellund.
It's main ideas:
- using IoC-container XML configuration as 'final' model (so, model interpretation is used instead of code-generation). This enables components composition and interaction from different DSM and using these components even outside of DSM space (from custom code, for instance)
- using XSL-based rules for describing transformations from XML-based DSM models to 'final' model
- designing DSM specifications in respect to existent DSM-s, transformations, components and infrastructure. This means that modeling should start from 'bottom' (low-level components, low level DSM-s) and level of 'abstraction' (in domain direction) should be decreased iteratively (so usually DSM will be transformed into more abstract DSM in this way: M1 -> M2 -> M3 -> ... -> M (ioc XML config - abstract and universal model representation).
This approach was successfully used for building such systems like comindwork.com and sqaservices.com extranet since early 2007. It still documented not very well but I'm going to describe it as part of my dissertation work and some aspects already covered by articles.
As part of dissertation work implementation I've started open-source project (NReco) that supports ideas of BU-DSM and makes them real (of course it contains only simple examples - for number of reasons I cannot share real-life infrastructure used for commercial system development).
Feel free to ask you questions, I'll be glad to read your feedbacks.
Views: 6
Add a Comment
-
Comment by Vitaliy Fedorchenko on March 25, 2009 at 11:48 -
Hi Mark, I hope BU-DSM will not disappoint you. Despite of the fact that it was born in cruel conditions of modern enterprise systems development, BU-DSM background refers to the theory of knowledge representation and ontology development. Will try to publish supporting materials as soon as possible (curse of priorities, you know).
-
Comment by Rui Curado on March 25, 2009 at 11:11 -
Mark, you know we can't please everyone and CG2009 is obviously no exception.
Let's not hijack Vitaliy's thread!
-
Comment by Mark Dalgarno on March 25, 2009 at 10:57
-
The focus of the Code Generation conferences is on the practical application of model-driven software development so many of the sessions are about lessons learned applying the approach in different problem domains.
We also have several presentations by sponsors but again these are about practical applications and I hope you'd agree that we do have represented many of the key commercial organisations in this field and we do ask sponsors not to make sales presentations.
In saying all that we do have several sessions looking at how to advance the state of the art, including for the first time a full-day workshop on day 1, a couple of think-tanks on software product lines and MDSD led by recognised researchers in the field and a couple of goldfish bowls where the level of discussion is set by the participants.
Please let me know what your other complaints are.
-
Comment by Rui Curado on March 25, 2009 at 10:39
-
I salute you to come forward and present a new approach to DSM/MDSD. This is something that should be encouraged, and is one of my complains about the CG2009 programme: No real innovative/disruptive/experimental presentations. Just a remix of what is already known, along with semi-commercial presentations...
© 2012 Created by Mark Dalgarno.
Powered by
You need to be a member of The Model Driven Software Network to add comments!
Join The Model Driven Software Network