| 1 |
On 10 September 2012 15:48, William Hubbs <williamh@g.o> wrote: |
| 2 |
> All, |
| 3 |
> |
| 4 |
> I have a regression in OpenRc wrt netplugd [1]. |
| 5 |
> |
| 6 |
> In researching this program, I have found that it and ifplugd, which is |
| 7 |
> the alternative, have been unmaintained for years. Also Debian has |
| 8 |
> declared netplugd to be obsolete in favor of ifplugd. |
| 9 |
|
| 10 |
The page referenced on the bug that says so appears to be talking |
| 11 |
about a different package than the one we have in the tree - they have |
| 12 |
different authors stated, and also, for the one we have the package is |
| 13 |
called netplug, with the executable called netplugd, whereas for the |
| 14 |
one declared obsolete the package itself is called netplugd. |
| 15 |
|
| 16 |
> Does anyone have any thoughts about whether we should keep OpenRC |
| 17 |
> support for one or both of these? |
| 18 |
|
| 19 |
There are a few options for this functionality (that I'm aware of): |
| 20 |
1) netplug: never used it so no particular comments. |
| 21 |
2) ifplugd: what I'm using now. I can't remember if there's a |
| 22 |
particularly good reason why I chose it, but I suspect it might have |
| 23 |
been for the audio feedback it provides when it detects a connection |
| 24 |
or disconnection. This probably isn't compelling enough by itself to |
| 25 |
keep the package if we'd otherwise want to remove it, but it is quite |
| 26 |
nice. |
| 27 |
3) wpa_supplicant: supposed to be able to do this even for wired |
| 28 |
interfaces, but I just did some experimenting and it seems broken - it |
| 29 |
thinks the cable is plugged in even when it isn't. |
| 30 |
4) dhcpcd: not sure when it was introduced, but current dhcpcd can |
| 31 |
detect when the link goes up and down, and request/renew its lease |
| 32 |
when it comes up. The only wrinkle that I can see here is that, if no |
| 33 |
ifplugd/netplug/wpa_supplicant is configured, OpenRC waits for it to |
| 34 |
receive a lease when starting the interface, rather than allowing it |
| 35 |
to background itself. |
| 36 |
|
| 37 |
So for dhcpcd, it might be enough to just make OpenRC aware that it |
| 38 |
doesn't need to wait for a lease when starting the interface. Keeping |
| 39 |
at least one of the other options working would still be required for |
| 40 |
other DHCP clients if they don't have similar functionality, or |
| 41 |
non-DHCP situations where it's necessary to do some sort of |
| 42 |
reconfiguration when the network is (dis)connected (such as OpenRC's |
| 43 |
arping module), assuming anyone cares about those of course. |
| 44 |
|
| 45 |
> |
| 46 |
> Thanks, |
| 47 |
> |
| 48 |
> William |
| 49 |
> |
| 50 |
> [1] https://bugs.gentoo.org/show_bug.cgi?id=427088 |
Archives