Create open source Tooling Framework?

[Geert Bellekens]

Following a discussion we (Paolo and me) had in this
topic we decided to start a new topic for this.

Here is what we discussed:

Secondly, some of our customers, especially new ones, don't want anything that isn't "mainstream" EA - and they generally view Add-ins as non-mainstream.  I think they'd view scripts as mainstream (hence my feature request for EXEC_SCRIPT)

I've encountered that sentiment before too, and I think it is wrong from a customer point of view.

It's the same for custom document generators, custom model validation, and I'm pretty sure for custom MDA transforms.

If you want any of those things to work in your environment you are going to spend quite a lot of time writing EA specific templates, scripts etc... I doubt whether the effort to get a decent RTF document template is less then the effort to create a document generator from scratch.

The advantages of a custom tool is that (given that you design it in a proper way )

• It will work exactly as you want it, for 100%, not more or less OK
  
• The effort is not lost when you would decide to switch CASE tools.
• You do not increase the (already strong) vendor lock-in.

It just seems hard to explain that to people who have "maximum out of the box" as one of their principles  
They should realize that creating templates for document generation, or MDA transforms is no "more" out of the box then creating a whole custom tool.

Geert

and

I agree, but it's hard to convince the customer.

Would it be easier if we had an (open source) modelling tooling framework that you can use to build your tools upon?

I've been thinking about something like that for a while, but I don't have the time (or money) to do something like that on my own.

The tools I'm writing today are all based on the same framework pattern, and they allow both variations in modelling method (style) as variations in the "engine" behind it.

I'm not sure if I could get the permission to open source my current (customer owned) work, but I could certainly replicate it.

Geert

So the idea is to build a tooling framework that could be used by different modelling methods and that could be ported to any UML CASE tool.

I've put some of my ideas around tooling into an article here
The pattern I use for my tooling on top of EA is described here
So far this pattern has served me well. It is based on my own experience with different customers that all had their own modelling method, and it even proved usefull in a CASE Tool migration scenario.

So what do you think? Would you use it? Would you contribute?

Geert

[Geert Bellekens]

Judging from the lack of response I gather there is not much interest

Geert

[marcelloh]

I would love to have some documentation generation that just fits like a glove (in every occasion, for every customer).

It looks promising, but I don't have time.

Keep me posted Geert, because the EA doc generator is far from being that glove.

[skiwi]

Would it be easier if we had an (open source) modelling tooling framework that you can use to build your tools upon?

There are few open source EA tools around, and certainly they require effort to use. http://www.enterprise-architecture.org/

We are not at the level of 'architecture maturity' where we would look at this yet,
and frankly, I'd personally be happy to be able to data model, diagram, and generate documentation more effectively.

EA has a great many features and power, but my observation is that there appears not to be a champion for subject areas, so, in the area I am currently working - data modelling, there are subject specific issues (eg the handling of foreign keys, or the flow through of attribute name changes to key names) or general issues (eg laying out models so that  lines don't cross objects, or giving lines more functionality, ie so they are not second class citizens), so these areas appear incomplete or inconsistent, or not polished and usable from a earth dwelling mere mortal's point of view.

[Geert Bellekens]

I think 'll start small, and begin with scratching my own itch(es).
One of those itches is an export (and then later import) functionality for the UmlCanvas diagrams.

I'll make sure to publish whatever gets developed, and then maybe others can jump on.

Geert

[salayande]

Geert,

You are doing work that should take modeling to the next level but most modelers (especially the average corporate Joe) are still struggling to obtain some value from EA software tools like Sparxsystem EA and many more out there. Open source projects are great for sharing expertise and so I support your suggestions.

At the same time, I suspect that ypu may be taking a business logic favoured by a lot of software companies to an extreme; that is, let the customer develop the product. I am not sure that a lot of business modelers out there are interested in knocking up their own solution, as that is the reason we pay maintenance fees for these features. We also work so hard modeling that by the time we get home, we are too tired to lift a finger.

Microsoft started a variant of this business model by delivering incomplete solution which we were all forced to test except that we all paid for the benefit of testing their products.

The next level to above approach is making customers extend software products in the name of Add-Ins or Extensions. This strategy rests on the belief that all customers are competent programmers and are able to write code to deliver product extensions. The problem with this assumption is that the core capabilities of the product that should be delivered are not and these extensions connot be delivered because the customers either lack the skills to create them or cannot identify a value proposition for them and so are not motivated to commit resources and time to their development.

The outcome of all this is that no particular customer segment is happy as features considered basic and standard in other tools are not delivered by the software publisher. The corporate modelers are most affected by this problem and hence a very slow uptake of a tool they have invested thieir time to learning. The result is potential market share loss to a more focused software publisher.

I have imagined a modeling tool that enables collaborative modeling with 3D type of modeling interface but I also recognise that many people may not be ready for this. We require the basics like seamless spreadsheet (visio) import-export, document management, and element level model versioning and many more.

I would like to participate in your project but I lack the programming skills and so cannot do much.

Segun