Re: Model repositories
I'm answering from the needs of a scientific study, but I don't know
the specific scope of ReMoDD:
Le 28-juin-06 à 15:23, Jean Bezivin a écrit :
> 1. Which kind of models do we want to collect?
> (UML models, UML "profiled" models? which profile, etc.)
We need to reflect industrial use and thus have all of them.
The profile description should be collected as well.
> 2. To which kind of metamodels should these models comply?
> UML 1.x, UML 2.0, UML 2.1, non UML metamodels?
The most recent the better, but we cannot be selective now and we
have to accept whatever industrial accept to contribute, which might
be much older.
> 3. What will be the storing format of these models?
> XMI 1.x, XMI 2.0, XMI 2.1, etc?
> 4. What will be the concrete repository?
> Implemented where? With which support?
> With which resources?
We have to use available resources, and coordinate existing projects.
> 5. What will be the tools available to
> feed or query this model repository?
Rather simple tools can already be quite effective:
an index and an ftp server would already do most of the work!
People will anyway download models and load them in their tools for
> 6. What will be the status of the contributed
> models? Open source? Which license? GPL? EPL?
It is enough that they can be freely analysed. The contributor can
keep all the copyrights.
> 7. What will be the accessibility to the deposited models?
> (This is related to the previous question, but several
> difficult situations may be found like some countries not
> allowed to access artefacts produced under industrial/military
> funding, etc.)
Control access of this type should not be part of the initial
We should first focus on free access.
> One general conclusion of this meeting was that there are
> already several initiatives that have been launched in Europe
> or North America on the subject of model repositories.
> As Jean-Marie suggests, it will be interesting to link
> these various initiatives. It is very unlikely that there will
> be one unique place for all these repositories, but a federated
> non-centralized model repository seems to make sense.
Yes, this is the way to go. I would recommend that data is
replicated: don't forget that projects have a short life and the
various projects should "pass the toked" to give the effect of a
permanent repository. For instance, a project that has no resources
to maintain its repository should give/lend its domain name and data
to a similar project.
Received on Thu 29 Jun 2006 - 08:41:56 BST