[arch-dev-public] adding a simple hostname binary to base (Was: [signoff] coreutils-8.12-3 initscripts-2011.06.4 net-tools-1.60-18)
Hi guys, Sorry to resurrect an old thread, but I think the context is relevant. On Fri, Jun 10, 2011 at 3:05 AM, Gaetan Bisson <bisson@archlinux.org> wrote:
This upgrade intends to fix the regression of coreutils' hostname not being as featureful as net-tools'.
Changes: - coreutils: revert to not building hostname - net-tools: revert to building hostname and dnsdomainname
At the moment we do not install net-tools as part of base, but some tools rely on hostname being present. I think it's presence is actually part of FHS ("hostname: Utility to show or set the system's host name"), so a reasonable assumption I suppose. To avoid having to install net-tools (or any binaries from it). I thought this might be a good compromise: Split out "coreutils-hostname" from coreutils. This new package should conflicts=(net-tools) and be added to base. Also, net-tools should be rebuilt to add provides=(coreutils-hostname). An important point is that we then will only get a very basic hostname utility which only does the minimum required to satisfy FHS (set/get hostname). If anything more "fancy" is needed such as "hostname -f" the legacy net-tools should be installed. Thoughts, Tom
On Tue, Aug 02, 2011 at 03:10:01PM +0200, Tom Gundersen wrote:
Hi guys,
Sorry to resurrect an old thread, but I think the context is relevant.
On Fri, Jun 10, 2011 at 3:05 AM, Gaetan Bisson <bisson@archlinux.org> wrote:
This upgrade intends to fix the regression of coreutils' hostname not being as featureful as net-tools'.
Changes: - coreutils: revert to not building hostname - net-tools: revert to building hostname and dnsdomainname
At the moment we do not install net-tools as part of base, but some tools rely on hostname being present. I think it's presence is actually part of FHS ("hostname: Utility to show or set the system's host name"), so a reasonable assumption I suppose.
To avoid having to install net-tools (or any binaries from it). I thought this might be a good compromise:
Split out "coreutils-hostname" from coreutils. This new package should conflicts=(net-tools) and be added to base. Also, net-tools should be rebuilt to add provides=(coreutils-hostname).
provides should be very general. 'hostname' would be more appropriate.
An important point is that we then will only get a very basic hostname utility which only does the minimum required to satisfy FHS (set/get hostname). If anything more "fancy" is needed such as "hostname -f" the legacy net-tools should be installed.
Thoughts,
We shouldn't make people install net-tools if the goal all along has been to move away from it. My vote is to split off the hostname binary from inetutils, which is just as full featured as hostname from net-tools. dave
On Tue, Aug 2, 2011 at 3:38 PM, Dave Reisner <d@falconindy.com> wrote:
On Tue, Aug 02, 2011 at 03:10:01PM +0200, Tom Gundersen wrote:
Split out "coreutils-hostname" from coreutils. This new package should conflicts=(net-tools) and be added to base. Also, net-tools should be rebuilt to add provides=(coreutils-hostname).
provides should be very general. 'hostname' would be more appropriate.
Good point. In that case, I guess both packages should provide 'hostname'.
We shouldn't make people install net-tools if the goal all along has been to move away from it. My vote is to split off the hostname binary from inetutils, which is just as full featured as hostname from net-tools.
That's probably even better then. It looks like inetutils is well maintained in core anyway, so no objections from me. Cheers, Tom
Am 02.08.2011 15:38, schrieb Dave Reisner:
We shouldn't make people install net-tools if the goal all along has been to move away from it. My vote is to split off the hostname binary from inetutils, which is just as full featured as hostname from net-tools.
All this seems like complete overkill. Why don't we just make people install inetutils? There are more tools from net-tools with no good replacement, like netstat.
On Tue, Aug 2, 2011 at 4:15 PM, Thomas Bächler <thomas@archlinux.org> wrote:
Am 02.08.2011 15:38, schrieb Dave Reisner:
We shouldn't make people install net-tools if the goal all along has been to move away from it. My vote is to split off the hostname binary from inetutils, which is just as full featured as hostname from net-tools.
All this seems like complete overkill. Why don't we just make people install inetutils?
There are more tools from net-tools with no good replacement, like netstat.
I guess we could add inetutils to base, and remove whatever binaries it provides from net-tools. Do we know that this would not lead to any regressions? Cheers, Tom
participants (3)
-
Dave Reisner
-
Thomas Bächler
-
Tom Gundersen