On Wed, Oct 10, 2018 at 3:15 PM Matthew Miller mattdm@fedoraproject.org wrote:
On Tue, Oct 09, 2018 at 06:07:44PM -0600, Al Stone wrote:
-- And the one question I have to add on to Christopher's wonderful list: I have a package where upstream releases about once a month, and each new release must by definition be backwards compatible (acpica-tools, specifically). I can think of no scenario where a module provides value to me or end-users; in fact, using anything other than the most recent causes problems. Do I have to create and maintain a module for this package anyway? Or are the defaults robust enough that a package can remain a package without touching modularity at all? The answer to this is completely unclear to me -- what I've read seems to imply that I must create a module definition regardless.
This actually seems like the ideal case for a single stream -- instead of maintaining rawhide, f29, f28, epel7, you'd just maintain "latest", and that would get build into all of these releases simultaneously.
What is the overhead of maintaining a module for a single package, plus the package itself, vs just maintaining the package the current way?
My understanding-- from skimming the documentation a few times and reading discussions about modularity-- is that I'd now need to keep track of two dist-git repositories, and two different metadata files. This feels like a lot of extra overhead. It also requires learning about a new thing-- modulemd files.
Is this really less work? I admit I haven't tried to do it myself yet, so I don't know. But part of the reason I haven't tried it is because I'm not sure if it will actually be better...
I guess it would be nice to read a sort of "modularity for the skeptical contributor" document or article that answers questions like this.
Ben Rosser