(This is about pkg/50075) I'm expanding to tech-pkg because this brings up a larger philosophical question beyond the dbus pkg. David Shao <davshao%gmail.com@localhost> writes: > The following patches enable the dbus daemon to even be started from > /etc/rc.conf on DragonFly 4.3-DEVELOPMENT and at least recent FreeBSD, > because the daemon is started using > > dbus_enable="YES" > > and not > > dbus="YES" > > An option enable-in-rcvar may be useful for other daemons on DragonFly > and FreeBSD. There are some assumptions in your statements above that I'd like to draw out, make more explicit, and discuss. Without loss of generality to Freebsd, I will just refer to DragonFly. You say that on DragonFly dbus "is started" by using dbus_enable in rc.conf rather than dbus. When using pkgsrc, one sets the rc.conf variable defines by the package, so one starts dbus by setting dbus=YES.. What I think you're saying is that some other packaging system uses dbus_enable instead. And further, that you want that same setting, intended for another system, to also control dbus in pkgsrc, when somehow you have done something to include pkgsrc's rc.d scripts in the system rc process. Which is a bit different than a simple "is started by" as a universal truth. In general, we have not tried to make pkgsrc use the same variables, users/groups, etc. as some other packaging system on various platforms. Instead, pkgsrc follows pkgsrc norms and is consistent across platforms. More or less as a consequence of this, we do not expect to share PREFIX or VARBASE with other packaging systems, because in general there will be conflicts. While your PR is only about dbus, it seems that if we were to adopt a notion that pkgsrc must mimic the rc variables, users/groups, paths, etc. of some other packaging system on particular platforms, we'd have to make a vast number of changes which would result in pkgsrc behaving differently on different platforms. While I can see that you might want that, I think others don't want it, because people use pkgsrc on multiple platforms to have a simpler and more regular approach. This note is only about the rcvar, but it seems likely to apply to the user/group and pathname issues. So I am curious if my assessment of the situation is accurate, or if you mean something else. Perhaps you can explain exactly what you're doing, and explain the details without making assumptions that follow from earlier decisions.
Attachment:
pgpWvUJcyFiVi.pgp
Description: PGP signature