On Friday, 8 May 2020 3:52:31 AM AEST Julien Cristau wrote: > That's not what "Provides" means though. What you're saying is > "systemctl provides a subset of the systemd package's functionality". > That's not good enough. Provides is much stronger than "there are cases > where this might be enough", and there's more to systemd than systemctl.
Basically what you are saying is that systemd provides so much that no package qualifies for "Provides: systemd". I have a metaphor in mind. Suppose systemd provides "house" with all its facilities such as microwave, dishwasher, fridge, washing machine, drier, etc. Re-implementing package have too much to provide so no package qualifies for "Provides: house" because under that doctrine providing package have to implement compatible facilities exactly the way systemd does so. I'm not sure what would be the incentive to re-write systemd with outcome closely resembling what systemd already provides... However realistic paradigm is more like when systemd provides "shelter". There are different types of _shelter_, e.g. motel, hotel, apartment and _tent_. The latter "Provides: shelter" but it needs no dishwasher and not expected to have it. Having it would require a lot of other facilities that _tent_ intentionally don't have (e.g. sewage, permanent water supply, 240V electricity, etc.). Note that feature-rich _hotel_ also don't provide dishwasher and also not expected to have it because other types of shelter are not meant to be identical to "house". Quite contrary they are meant to be different. Based on that analogy I think "Provides: systemd" is acceptable if we don't take "systemd" literally as the collection of all systemd-specific features/ facilities/interfaces altogether. Just like we would not consider _library_ to be a _library_ only if it provides a particular and very specific set of books similar to a certain public library. > Also, per policy §3.6, virtual packages outside the defined agreed-upon > names should only be used "amongst a cooperating group of packages". It > seems clear to me this is neither of those cases. Perhaps you know better but I'm not sure how much this matters in the context of the intended outcome. > [...] or the > php-fpm maintainer to forgo using functionality that is only available > in systemd, but abusing package relationships the way you're doing now > is just not on. Noted. I disagree with strong statement about "abusing package relationships" but agree with you that removing "Provides: systemd" from "systemctl" package would be desirable as long as "php-fpm" package makes changes to allow our packages to be co-installable. I am making attempt to convince Ondřej to make changes: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=959174#52 -- Regards, Dmitry Smirnov. --- No system of mass surveillance has existed in any society that we know of to this point that has not been abused. -- Edward Snowden
signature.asc
Description: This is a digitally signed message part.
