Comments on Moserware: Towards Moore's Law Software: Part 2 of 3

Juha-Pekka Tolvanen: Interesting thought -- I hadn...

2008-04-18T22:31:00.000-04:00

Juha-Pekka Tolvanen: Interesting thought -- I hadn't considered that companies might be actively trying not to let the fact that their using these concepts for fear they'd lose an advantage.

When it comes to adoption, I see that there is no ...

2008-04-18T04:46:00.000-04:00

When it comes to adoption, I see that there is no single reason holding the domain-specific (modeling) languages from becoming mainstream. One obvious reason why we don’t hear much about success stories on using domain-specific languages and models is that they are often propriety and contained within one company only. For instance, over 90% of our customers don’t allow us to mention details regarding their DSM activities. And I understand this very well -- if you are 10x faster than your competitors why you would like to reveal your advantage?

Peter Christensen: Paul Beel's group sounds intere...

2008-04-16T21:11:00.000-04:00

Peter Christensen: Paul Beel's group sounds interesting and I checked out his blog. Unfortuntaely, the 100 minute drive would make it a bit too far to attend regularly.

Thanks for the contact info though, never know when our paths might meet.

Juha-Pekka Tolvanen: Wow! Impressive experience, especially with the paper. I agree with the points you made.

I have a follow-up question though: in your opinion, based off all the work you've done, what do you think is holding these tools and ideas from becoming more mainstream? Are they too niche and require too much training or customization?

I see that one sort of domain-specific models is t...

2008-04-16T15:11:00.000-04:00

I see that one sort of domain-specific models is those based on flow type, like user navigation and interaction, task workflow or various services, like call processing services described in detail in: http://www.metacase.com/support/45/manuals/Call%20Processing%20Language%20Example.pdf

Another soft of models is those describing static structures, like automotive architectures in AUTOSAR (http://www.metacase.com/images/SoftwareComponentDiagram.png) or insurance products (http://www.metacase.com/images/insurance.gif).

If behavior is a relevant part of the domain and there is need for a more comprehensive code generation, then various state machines are preferable models of computation.

Often in practice there is a need to combine multiple of languages. Their concrete syntax does not need to be graphical but can also be based on matrix and table presentations along with some text. These possibilities are naturally dependent on the tool. We wrote few years ago a paper on various language structures and how companies tend to build domain-specific languages. In this paper we analyzed 20+ industry cases. If interested, you may find the paper here: http://users.jyu.fi/~jpt/TolKelSPLC2005.pdf

When it comes to Intentional Software I see several similarities, but I have only seen one of their case presented at OOP last January (http://www.metacase.com/blogs/jpt/blogView?showComments=true&entry=3378978081). That talk mostly gave the rationale for raising the level of abstraction and creating domain-specific languages than discussed how the tool works in practice. Issues like how the languages are defined and maintained or how they evolve along with models created already created were not unfortunately addressed.

Jeff, you should get in touch with Paul Beel (paul...

2008-04-16T13:30:00.000-04:00

Jeff, you should get in touch with Paul Beel (paul dot beel at gmail dot com) and check out his blog at http://novacode.wordpress.com/ . He has started a group in Terra Haute called the Indiana Programming Meetup. Their first meeting was about Common Lisp (http://novacode.wordpress.com/2008/03/26/meeting-main-topic-common-lisp/ ) and they had the guys from Paragent come and show off their system and CUSP, the CL plugin for Eclipse. It's not exclusively Lisp (next meeting is by a PHP guy) but might be up your alley.

-Peter

Peter: Thanks for the encouragement. Hopefully par...

2008-04-15T22:45:00.000-04:00

Peter: Thanks for the encouragement. Hopefully part 3 lives up to your expectations :)

I checked out your blog and saw you mention the Chicago Lisp meeting. That definitely sounds interesting. It'd be neat to have something similar in Indianapolis.

Juha-Pekka: Thanks for the link! I checked out the cases. Do you see DSM mostly being used for workflow type scenarios? It seemed like MetaEdit+ is the popular tool to use. Have you seen ones built with Simonyi's tools?

Nice post. You might be interested to see industry...

2008-04-15T17:52:00.000-04:00

Nice post. You might be interested to see industry cases on creating and using domain-specific (modeling) languages at http:://www.dsmforum.org/cases.html

Jeff, you've saved me the trouble of writing somet...

2008-04-15T11:25:00.000-04:00

Jeff, you've saved me the trouble of writing something that has been on my mind for a while. I can't wait for Part 3!