We were discussing about that.My point is that the system should work if the catalog fall apart.
Now others told me that in other community this is the inverse. People only refer to the catalog.
Since we did not get last year and this year the engineer I requested to work on the validation of distirbution we did not make progress
on that front because we are busy.
For me I do not use SqueakSource not because it is written Squeak but because I think that squeaksource is over and too old.
Stef
I'm currently browsing the various ConfigurationOf in MetaRepoForPharo30 catalog hoping to find a pattern to serve as guidelines, but what I see is a messy variety of solutions:
- direct reference to specific source repositories of package B as Johan suggested (and the successive baselines are effectively tracking motions of B)
IMO the worst thing, I don't manage B and don't want to update my ConfigurationOfA each time B moves.
- references to ConfigurationOfB thru squeaksource MetacelloRepository.
My preferred solution. I don't care if the catalog is centralized on squeaksource, Smalltalkhub or whatever, but I want a centralized catalog
Me too.
Now the question is the following one.
Ideally I would like to have one inbox and that configuration move to the catalog for their version.
(Because I do not see the point of having around old versions of projects that do not load in the given version).
I explained that in the Pharo vision document.
Now may be other solutions are better.
Stef
- refrences thru MetaRepo catalog du jour (tracking motions of MetaRepo across baselines)
My original question, using secondary catalogs is not a problem if they are just a duplicata, but I prefer to ask, because it sounds strange to have multiple catalogs just to implement a filter (the filter being select those packages that are tested and approved in pharo x.y)... Certainly a workaround for a missing feature of Metacello.
- a mixture of the 3 solutions above for some packages!
That appears as randomness from my POV, or like of guidance
So again, what are the best practices currently, and could we think of it in a bit longer term?