[furcas-discussion] Where are things?

[Stephan Erb]

Hi Axel,

this is my 3rd attempt to write this mail. I'll try to keep it short
this time, so that the actual message doesn't get buried :-)

> I wonder where things are with FURCAS right 
> now. Is our Bugzilla our "backlog?" Any other sources where we can know 
> what the big open issues are?
Backlog is badly scattered (wiki, bugzilla, loose paper floating around
my desk...). Where shall we put things?

> I still feel that the refactoring that I started which was intended to 
> remove the DelayedReference... stuff still hasn't completed, or has 
> anyone continued with that?
Nothing changed in that regard.

During refactoring in the master branch, I inspected several pieces of
code dealing with delayed references and concluded that the scope
handling and the dependency injection paths have to be reviewed. The way
editor works does not seem to fit the interface of the model updaters /
reference resolving infrastructure.

For example, when a model updater kicks in, it has to work on the exact
scope of the editor and also know about transient model elements in the
internal resource. The model updater should not source the internal
resource of an editor that is parsed concurrently in another background
thread.

> Where are things regarding the pretty-printer? What about the IMP 
> migration? Do we have usable editors? Does the wizard work?
Pretty Printer has seen some minor bugfixes, but it still rather
problematic. A whiteboard session to clarify its requirements and to
understand its current shortcomings seems worthwhile (e.g., re-print a
model element when it has been modified within another editor).

There development of IMP editors is pretty much on-hold
until the incremental parser is fixed. One cannot really test the editor
because the parser dies after a few keystrokes.

I believe that the editors will be usable once the parser is fixed (bold
assumption).


> I have three potential use cases.
> [...]
> What do you think? Is FURCAS mature enough for any of these three use cases?

Yes and no. I don't think that we are mature enough. However,
approaching usecases is the only chance we have. Without a closed
feedback loop we are pretty much developing in the dark and might end
nowhere.

Having concrete uses cases (e.g., runlet) seems to have served FURCAS
really well in the past. 

Minimal required bugfixes that come to mind:
      * Incremental parser fixes. This includes the 8 broken testcases,
        but there might also be more bugs.
      * Prevent XMI/XML resource parser crashes during builds in the
        target workbench. (Sebastian is on it)

More advanced topics that we should have on our agenda:
      * the whole story of delayed reference resolving and model
        updaters
      * a better incremental pretty printer


Axel, have you considered re-activating the NGPM syntax? Maybe that is a
good start for you to get a feel for the capabilities of the new
infrastructure. 

Kind Regards,
Stephan