On 03/24/2017 09:58 AM, Alberto Molina Coballes wrote: > 2017-03-24 1:26 GMT+01:00 Thomas Goirand <z...@debian.org > <mailto:z...@debian.org>>: > > > Hi, > > Unfortunately, not everyone wants to use IPv6. For those who don't, > imposing radvd as a hard dependency is a bad idea. IMO, it's best to let > every user choose what it wants to do to match its use case, and > manually install radvd. > > > Hi Thomas, > > There's no reference about the need to install radvd to configure > neutron with IPv6 support in the documentation and IMO is confusing when > a configuration option fails because a necessary package is not > installed. radvd is a small package with no additional dependencies, so > I don't think is bad idea to include it as a dependency.
The problem isn't the size of the package or its eventual dependencies. The issues is that we want to allow to *not* install or run radvd if a user doesn't want it. Indeed, it may destroy an already running ipv6 setup, or add a potentially not necessary complexity and security surface of attack that one may now want. > We could add radvd as Suggests: though, but that is too late to ask for > a migration exception, so this change wont be in Stretch. > > Yes, I agree with you, it's too late for Stretch but maybe it could be > included in the next release. Will do. Thomas Goirand (zigo)
