Vasco Amaral, Univ. of Mannheim - PowerPoint PPT Presentation

About This Presentation
Title:

Vasco Amaral, Univ. of Mannheim

Description:

Steve Cook, Microsoft. Daniel Dvorak, JPL/Cal Tech. Lars Grunske, HPI. James Tolles, Earthlink. Jon Whittle, NASA AMES Research Center. DSM Practice ... – PowerPoint PPT presentation

Number of Views:33
Avg rating:3.0/5.0
Slides: 9
Provided by: juhapekkat
Learn more at: http://www.dsmforum.org
Category:
Tags: amaral | cook | james | mannheim | univ | vasco

less

Transcript and Presenter's Notes

Title: Vasco Amaral, Univ. of Mannheim


1
DSM Practice
  • Vasco Amaral, Univ. of Mannheim
  • Steve Cook, Microsoft
  • Daniel Dvorak, JPL/Cal Tech
  • Lars Grunske, HPI
  • James Tolles, Earthlink
  • Jon Whittle, NASA AMES Research Center

2
DSMs vs DSLs
  • A Domain Specific (Modeling) Language is a
    language designed to express the variabilities
    within a problem space
  • Metamodel abstract syntax concepts
  • Notations (text and diagrams)
  • Targets mapping to the targets
  • A Domain Specific Model is a sentence in a
    Domain Specific Modeling Language

3
Creating a DSL
  • The metamodel for a DSL is like an object model
    of the problem domain
  • Somewhat like good old OOAD underline the
    nouns
  • Also captures relationships, well-formedness
    rules, and notations

4
DSLs vs Frameworks
  • Frameworks define hotspots of variability
  • DSLs allow you to specify values for the
    variables
  • Frameworks DSLs Systems

5
When are DSLs good?
  • A family of problems where the variation is large
    enough to be non-trivial to do by hand, but not
    too large
  • Space missions (Kalman filters, control systems,
    telemetry)
  • High energy physics (data mining, querying,
    visualization)
  • Video card ltPAL, NTSC, SECAMgt
  • Family of architectures with specific
    non-functional properties
  • Organization chart
  • Venn diagrams
  • What software problems are not best solved with a
    domain specific language?
  • You need a family of problems with identified
    variabilities and to design a language that
    allows you to assign values to the variables
  • The economics must work

6
Cost vs Benefit
  • How many uses to make it economically viable to
    do domain engineering?
  • Size of system
  • Programmer/tester time saved
  • Time during which the program is useful
  • Competitive advantage/time to market
  • Number of similar systems
  • Product family experiences show that if there are
    more than 3 variants to be developed, DSM is
    cost-justified
  • Others?

7
Organisational issues of DSM introduction
  • In some organisations, existing developers want
    to just write the code like they have for the
    last N years, are not interested and dont see
    the value
  • Prototype to show them the benefits
  • Identify an agent of change in the organisation
  • Workshops, conferences, books, ecosystems
  • Corporate new technology group responding to
    management pressure to increase IT efficiency
  • Commissioned RD in DSM techniques

8
Hot topics
  • Does DSModeling in some way represent a decrease
    in interest in objects?
  • How to validate models?
  • Invariants, consistency checking
  • How to generate tests as well as code?
  • How to ensure a good user experience from a
    generated tool?
  • How to separate domain elements from non-domain
    elements?
  • How to do mappings? One-way, both ways
Write a Comment
User Comments (0)
About PowerShow.com