[arch-projects] hostname regression
Hi Gaetan, I just thought of this possible solution: "hostname" does not really have to be in core, except for its use in initscripts (correct me if I'm wrong here), so it is ok to keep it in net-tools, and not install net-tools by default (assuming initscripts somehow can set the hostname on boot). If that's correct, we could just install the hostname from coreutil in some place that is not in PATH, such as /lib/initscripts/hostname (without installing the manpage), and keep the old hostname in net-tools. This is similar to what you suggested (putting it in /usr), except that it is a bit more clear: either you have hostname or you don't, you don't have different versions of it depending on what packages are installed. -t
On Thu, Jun 9, 2011 at 1:05 PM, Tom Gundersen <teg@jklm.no> wrote:
Hi Gaetan,
I just thought of this possible solution: "hostname" does not really have to be in core, except for its use in initscripts (correct me if I'm wrong here), so it is ok to keep it in net-tools, and not install net-tools by default (assuming initscripts somehow can set the hostname on boot).
If that's correct, we could just install the hostname from coreutil in some place that is not in PATH, such as /lib/initscripts/hostname (without installing the manpage), and keep the old hostname in net-tools.
This is similar to what you suggested (putting it in /usr), except that it is a bit more clear: either you have hostname or you don't, you don't have different versions of it depending on what packages are installed.
-t
Also, inetutils also could provide a hostname and even an ifconfig binary. They are currently not included in the package as they are (or were) provided by other packages. See the list of utilities that we could add in inetutils: --disable-tftp --disable-tftpd \ --disable-ping --disable-ping6 \ --disable-logger --disable-syslogd \ --disable-inetd --disable-whois \ --disable-uucpd --disable-hostname \ --disable-ifconfig --disable-traceroute I don't know how they compare to the other versions of these tools provided by net-tools, coreutils, etc. But they might be a good alternative. Eric
Hi Eric, On Thu, Jun 9, 2011 at 10:06 PM, Eric Bélanger <snowmaniscool@gmail.com> wrote:
Also, inetutils also could provide a hostname and even an ifconfig binary. They are currently not included in the package as they are (or were) provided by other packages. See the list of utilities that we could add in inetutils:
--disable-tftp --disable-tftpd \ --disable-ping --disable-ping6 \ --disable-logger --disable-syslogd \ --disable-inetd --disable-whois \ --disable-uucpd --disable-hostname \ --disable-ifconfig --disable-traceroute
I don't know how they compare to the other versions of these tools provided by net-tools, coreutils, etc. But they might be a good alternative.
That's probably a very good idea. We'd just have to check the functionality versus net-tools. The question remains: what to do with initscripts. It just needs to set the hostname, seems silly to pull in net-tools or inetutils just for this...
On Thu, Jun 9, 2011 at 4:15 PM, Tom Gundersen <teg@jklm.no> wrote:
Hi Eric,
On Thu, Jun 9, 2011 at 10:06 PM, Eric Bélanger <snowmaniscool@gmail.com> wrote:
Also, inetutils also could provide a hostname and even an ifconfig binary. They are currently not included in the package as they are (or were) provided by other packages. See the list of utilities that we could add in inetutils:
--disable-tftp --disable-tftpd \ --disable-ping --disable-ping6 \ --disable-logger --disable-syslogd \ --disable-inetd --disable-whois \ --disable-uucpd --disable-hostname \ --disable-ifconfig --disable-traceroute
I don't know how they compare to the other versions of these tools provided by net-tools, coreutils, etc. But they might be a good alternative.
That's probably a very good idea. We'd just have to check the functionality versus net-tools.
The question remains: what to do with initscripts. It just needs to set the hostname, seems silly to pull in net-tools or inetutils just for this...
What you suggested above seems right: have initscripts provide it's copy of hostname. The package is already arch dependent because of the minilogd binary so adding another binary doesn't change anything. And at 22KB, hostname is quite small so we don't wast much of diskspace. The other solution is what we do now: have hostname in a base package that everyone has installed like coreutils.
On Thu, Jun 9, 2011 at 10:38 PM, Eric Bélanger <snowmaniscool@gmail.com> wrote:
On Thu, Jun 9, 2011 at 4:15 PM, Tom Gundersen <teg@jklm.no> wrote:
Hi Eric,
On Thu, Jun 9, 2011 at 10:06 PM, Eric Bélanger <snowmaniscool@gmail.com> wrote:
Also, inetutils also could provide a hostname and even an ifconfig binary. They are currently not included in the package as they are (or were) provided by other packages. See the list of utilities that we could add in inetutils:
--disable-tftp --disable-tftpd \ --disable-ping --disable-ping6 \ --disable-logger --disable-syslogd \ --disable-inetd --disable-whois \ --disable-uucpd --disable-hostname \ --disable-ifconfig --disable-traceroute
I don't know how they compare to the other versions of these tools provided by net-tools, coreutils, etc. But they might be a good alternative.
That's probably a very good idea. We'd just have to check the functionality versus net-tools.
The question remains: what to do with initscripts. It just needs to set the hostname, seems silly to pull in net-tools or inetutils just for this...
What you suggested above seems right: have initscripts provide it's copy of hostname. The package is already arch dependent because of the minilogd binary so adding another binary doesn't change anything. And at 22KB, hostname is quite small so we don't wast much of diskspace.
The other solution is what we do now: have hostname in a base package that everyone has installed like coreutils.
Hm, ok I'll go back to my original suggestion then. Have a hostname just for initscripts and let people install net-tool (or inetutils if that's better) to get the normal functionality back. Though the initscripts specific hostname should still be shipped with coreutils I think, just moved to /lib/initscripts, so only initscripts will know about it ;-) That makes packaging easier (I think). -t
Am 09.06.2011 22:15, schrieb Tom Gundersen:
The question remains: what to do with initscripts. It just needs to set the hostname, seems silly to pull in net-tools or inetutils just for this...
Why would it be silly? Anyway: echo $HOSTNAME > /proc/sys/kernel/hostname No external tools required.
On Thu, Jun 9, 2011 at 2:53 PM, Thomas Bächler <thomas@archlinux.org> wrote:
Am 09.06.2011 22:15, schrieb Tom Gundersen:
The question remains: what to do with initscripts. It just needs to set the hostname, seems silly to pull in net-tools or inetutils just for this...
Why would it be silly? Anyway:
echo $HOSTNAME > /proc/sys/kernel/hostname
No external tools required.
Thanks, I was about to add that. as far as I know that is all setting the hostname does. Also, it would be nice if the hostname command maintained backwards compatibility, a number of my scripts just broke because hostname no longer has the -f option. -Thomas S Hatch
On Thu, Jun 9, 2011 at 10:53 PM, Thomas Bächler <thomas@archlinux.org> wrote:
Anyway:
echo $HOSTNAME > /proc/sys/kernel/hostname
No external tools required.
D'oh! Problem solved then. I was thinking earlier "surely this must be possible to do directly", but I rejected it. In that case, this is no longer any of my business, and we can just simple revert to putting hostname in net-tools for the time being, and possibly move it to inetutils in the future (but I won't be fighting that battle ;-) ). -t
[2011-06-09 22:58:49 +0200] Tom Gundersen:
In that case, this is no longer any of my business, and we can just simple revert to putting hostname in net-tools for the time being
Very well, then. Let me know when the updated initscripts hit [testing] and I'll put updated packages for coreutils (without hostname) and net-tools (with hostname, and updated install message). Cheers. -- Gaetan
Am 09.06.2011 22:06, schrieb Eric Bélanger:
Also, inetutils also could provide a hostname and even an ifconfig binary. They are currently not included in the package as they are (or were) provided by other packages. See the list of utilities that we could add in inetutils:
--disable-tftp --disable-tftpd \ --disable-ping --disable-ping6 \ --disable-logger --disable-syslogd \ --disable-inetd --disable-whois \ --disable-uucpd --disable-hostname \ --disable-ifconfig --disable-traceroute
I don't know how they compare to the other versions of these tools provided by net-tools, coreutils, etc. But they might be a good alternative.
Does it have netstat? With a proper netstat replacement, we could get rid of net-tools.
[2011-06-09 22:52:39 +0200] Thomas Bächler:
Does it have netstat? With a proper netstat replacement, we could get rid of net-tools.
iproute2 provides the ss command. -- Gaetan
Am 09.06.2011 22:59, schrieb Gaetan Bisson:
[2011-06-09 22:52:39 +0200] Thomas Bächler:
Does it have netstat? With a proper netstat replacement, we could get rid of net-tools.
iproute2 provides the ss command.
netstat is a standard unix tool with (more or less) standardized syntax and output. ss has all the features, but is not the same.
participants (5)
-
Eric Bélanger
-
Gaetan Bisson
-
Thomas Bächler
-
Thomas S Hatch
-
Tom Gundersen