Re: [arch-general] [arch-dev-public] iproute2 to base
On Tue, 16 Oct 2012 10:51:34 +0200 Thomas Bächler <thomas@archlinux.org> wrote:
I am fine with add wpa_supplicant, iw and crda to base if we really want wireless support installed by default.
Not that I install many systems, once installed... But I for one would appreciate having wireless support on the install media. -- Joakim
Am 16.10.2012 14:05, schrieb Joakim Hernberg:
On Tue, 16 Oct 2012 10:51:34 +0200 Thomas Bächler <thomas@archlinux.org> wrote:
I am fine with add wpa_supplicant, iw and crda to base if we really want wireless support installed by default.
Not that I install many systems, once installed... But I for one would appreciate having wireless support on the install media.
The install media already have wireless support. The 'base' group is about the default set of packages that are installed when running # pacstrap /mnt base Right now, this pull in these packages: https://www.archlinux.org/groups/x86_64/base/ (only the one marked "Core"). I am really interested in opinions as to what we should provide here. I think that a user without wired access could write # pacstrap /mnt base wpa_supplicant wpa_actiond iw crda and get full automatic wireless support on the installed system via netcfg.
On Tue, 16 Oct 2012 15:22:05 +0200 Thomas Bächler <thomas@archlinux.org> wrote:
Am 16.10.2012 14:05, schrieb Joakim Hernberg:
On Tue, 16 Oct 2012 10:51:34 +0200 Thomas Bächler <thomas@archlinux.org> wrote:
I am fine with add wpa_supplicant, iw and crda to base if we really want wireless support installed by default.
Not that I install many systems, once installed... But I for one would appreciate having wireless support on the install media.
The install media already have wireless support. The 'base' group is about the default set of packages that are installed when running # pacstrap /mnt base Right now, this pull in these packages: https://www.archlinux.org/groups/x86_64/base/ (only the one marked "Core").
I am really interested in opinions as to what we should provide here. I think that a user without wired access could write # pacstrap /mnt base wpa_supplicant wpa_actiond iw crda and get full automatic wireless support on the installed system via netcfg.
I fully support having netcfg in base (and as a default network backend in arch) because it is far better than the alternatives :) I don't think that wpa_supplicant/crda belongs in base (for instance routers don't need wpa_supplicant but may require hostapd), but iw (and iproute2) definitely has to go there as it provides some hardware management capabilities. Just an opinion... -- Leonid Isaev GnuPG key: 0x164B5A6D Fingerprint: C0DF 20D0 C075 C3F1 E1BE 775A A7AE F6CB 164B 5A6D
[2012-10-16 10:41:09 -0500] Leonid Isaev:
I fully support having netcfg in base (and as a default network backend in arch) because it is far better than the alternatives :) I don't think that wpa_supplicant/crda belongs in base (for instance routers don't need wpa_supplicant but may require hostapd), but iw (and iproute2) definitely has to go there as it provides some hardware management capabilities.
Since routers do not need netcfg any more than they do wpa_supplicant, with your reasoning, it should not be in base either... If we stick to the definition that the base group should contain everything needed too boot up a minimal system and connect it to the network, then I do not see how you can consider wpa_supplicant optional. -- Gaetan
On 10/16/2012 08:21 PM, Gaetan Bisson wrote:
[2012-10-16 10:41:09 -0500] Leonid Isaev:
I fully support having netcfg in base (and as a default network backend in arch) because it is far better than the alternatives :) I don't think that wpa_supplicant/crda belongs in base (for instance routers don't need wpa_supplicant but may require hostapd), but iw (and iproute2) definitely has to go there as it provides some hardware management capabilities.
Since routers do not need netcfg any more than they do wpa_supplicant, with your reasoning, it should not be in base either...
YMMV apparently, but in my experience a router needs: (1) Some way to stick to the (usually creepy) ISP DHCP server, i.e. keep retrying to obtain IP if the DHCP server doesn't respond. (2) Bridging support. The former is solved with net-auto-wired (ifplugd is quite good), while the latter -- with "bridge" profiles in netcfg. So without netcfg I would have to write my own boot scripts.
If we stick to the definition that the base group should contain everything needed too boot up a minimal system and connect it to the network, then I do not see how you can consider wpa_supplicant optional.
I understand your logic, but still think that wpa_supplicant should be optional. Since there are no core images, anyone who wants to use a machine as a station will install wpa_supplicant anyways over the already working network...
Leonid Isaev wrote:
On 10/16/2012 08:21 PM, Gaetan Bisson wrote:
[2012-10-16 10:41:09 -0500] Leonid Isaev:
I fully support having netcfg in base (and as a default network backend in arch) because it is far better than the alternatives :) I don't think that wpa_supplicant/crda belongs in base (for instance routers don't need wpa_supplicant but may require hostapd), but iw (and iproute2) definitely has to go there as it provides some hardware management capabilities.
Since routers do not need netcfg any more than they do wpa_supplicant, with your reasoning, it should not be in base either...
YMMV apparently, but in my experience a router needs: (1) Some way to stick to the (usually creepy) ISP DHCP server, i.e. keep retrying to obtain IP if the DHCP server doesn't respond. (2) Bridging support.
The former is solved with net-auto-wired (ifplugd is quite good), while the latter -- with "bridge" profiles in netcfg. So without netcfg I would have to write my own boot scripts.
If we stick to the definition that the base group should contain everything needed too boot up a minimal system and connect it to the network, then I do not see how you can consider wpa_supplicant optional.
I understand your logic, but still think that wpa_supplicant should be optional. Since there are no core images, anyone who wants to use a machine as a station will install wpa_supplicant anyways over the already working network...
The idea at some point [0] was to make base-{networking,wireless-networking} groups. I dont know if that can be considered today, but this is more or less the idea of why i had requested for wpa_supplicant to leave base [1] and why i have requested the same for ppp [2]. If this old concept, or a similar one is implemented personally even though wpa_supplicant is essential for me as well, i see no reason to have any of those in base. On the other hand by having netcfg in the base group you essentially provide support for both. BTW i dont understand the reason why the base group has anything to do with archiso, since archiso adds and will probably be always adding packages on top of the base group from all the other repos in order to provide additional functionality. Argumenting that foo should be in base cause we want support for it in the install media is no argument at all. [0]: https://bugs.archlinux.org/task/12890 [1]: https://bugs.archlinux.org/task/22482 [2]: https://bugs.archlinux.org/task/22480
On Fri, Oct 19, 2012 at 03:10:15AM +0300, Greg Bouzakis wrote:
Leonid Isaev wrote:
On 10/16/2012 08:21 PM, Gaetan Bisson wrote:
[2012-10-16 10:41:09 -0500] Leonid Isaev:
I fully support having netcfg in base (and as a default network backend in arch) because it is far better than the alternatives :) I don't think that wpa_supplicant/crda belongs in base (for instance routers don't need wpa_supplicant but may require hostapd), but iw (and iproute2) definitely has to go there as it provides some hardware management capabilities.
Since routers do not need netcfg any more than they do wpa_supplicant, with your reasoning, it should not be in base either...
YMMV apparently, but in my experience a router needs: (1) Some way to stick to the (usually creepy) ISP DHCP server, i.e. keep retrying to obtain IP if the DHCP server doesn't respond. (2) Bridging support.
The former is solved with net-auto-wired (ifplugd is quite good), while the latter -- with "bridge" profiles in netcfg. So without netcfg I would have to write my own boot scripts.
If we stick to the definition that the base group should contain everything needed too boot up a minimal system and connect it to the network, then I do not see how you can consider wpa_supplicant optional.
I understand your logic, but still think that wpa_supplicant should be optional. Since there are no core images, anyone who wants to use a machine as a station will install wpa_supplicant anyways over the already working network...
The idea at some point [0] was to make base-{networking,wireless-networking} groups. I dont know if that can be considered today, but this is more or less the idea of why i had requested for wpa_supplicant to leave base [1] and why i have requested the same for ppp [2]. If this old concept, or a similar one is implemented personally even though wpa_supplicant is essential for me as well, i see no reason to have any of those in base. On the other hand by having netcfg in the base group you essentially provide support for both.
BTW i dont understand the reason why the base group has anything to do with archiso, since archiso adds and will probably be always adding packages on top of the base group from all the other repos in order to provide additional functionality. Argumenting that foo should be in base cause we want support for it in the install media is no argument at all.
[0]: https://bugs.archlinux.org/task/12890 [1]: https://bugs.archlinux.org/task/22482 [2]: https://bugs.archlinux.org/task/22480
IMHO, wpa_supplicant should no more be in base than should linux-atm, brltty, or any of the firmware packages, e.g. ipw2100-fw. In essence, it's a package for the support of certain hardware / network configurations (albeit, widely used ones) that aren't handled directly by the kernel. -- David J. Haines djhaines@gmx.com
[2012-10-18 15:15:02 -0400] Leonid Isaev:
On 10/16/2012 08:21 PM, Gaetan Bisson wrote:
Since routers do not need netcfg any more than they do wpa_supplicant, with your reasoning, it should not be in base either...
YMMV apparently, but in my experience a router needs: (1) Some way to stick to the (usually creepy) ISP DHCP server, i.e. keep retrying to obtain IP if the DHCP server doesn't respond. (2) Bridging support.
The former is solved with net-auto-wired (ifplugd is quite good), while the latter -- with "bridge" profiles in netcfg. So without netcfg I would have to write my own boot scripts.
You may solve these problems with whatever piece of software you want. On my router, I simply use iptables' FORWARD chain. My point being that there are dozens of apps that do the same thing netcfg does; however wpa_supplicant is the only one that does what it does.
I understand your logic, but still think that wpa_supplicant should be optional. Since there are no core images, anyone who wants to use a machine as a station will install wpa_supplicant anyways over the already working network...
And how is this not different for netcfg? I understand you have use for one but not the other, but please try to be a little objective... -- Gaetan
participants (6)
-
David J. Haines
-
Gaetan Bisson
-
Greg Bouzakis
-
Joakim Hernberg
-
Leonid Isaev
-
Thomas Bächler