On 2016-05-25 18:00:55, Joachim Breitner wrote: > Hi, > > Am Mittwoch, den 25.05.2016, 09:34 +0200 schrieb Raphael Hertzog: > > I fear that adding such expressivity is encouraging bad practice. While > > I can understand that a single repository can be convenient and easier > > to manage than proper "mr" usage, it also feels wrong on many levels: > > - you don't know what branch is relevant for what package > > - you are almost forced to not inject upstream sources to avoid excessive > > growth of the single repository > > - you have to namespace all your tags (and branches) > > > > In general, this choice goes against the various recommendations we > > tried to define in DEP 14: http://dep.debian.net/deps/dep14/ > > for some value of "we". The choice not to include the upstream sources > was integral part of the design of this repository layout, and not a > by-product. Namespaced tags are not necessarily a disadvantage. And > that branching affects all packages together could also be seen as an > advantage. > > I am not advocating this as a best practice in the general case, but > when our case where the packaging of each individual package is > trivial, and it is the orchestration of a large number of packages that > is our main concern, the usually recommended workflows are not ideal > and something like this works better. I think the mismatch in the discussion here is that Raphael thinks of individual packages, whereas we rather talk about the "unified Haskell packaging" effort. Its output is individual src/bin debs, but conceptually it's a single, large entity that is (in theory at least) self-consistent (w.r.t. package-plan, internal dependencies are fulfilled, etc.). So we can't really compare DHG_packages with dep14 and normal packages. As such, with all respects to Raphael, I think we're not talking about the same thing. regards, iustin
Attachment:
signature.asc
Description: PGP signature