Gentoo Archives: gentoo-dev

From: Rich Freeman <rich0@g.o>
To: gentoo-dev <gentoo-dev@l.g.o>
Cc: Thomas Sachau <tommy@g.o>
Subject: Re: [gentoo-dev] newsitem: initramfs required on Linux systems with separate /usr
Date: Wed, 25 Sep 2013 11:20:54
Message-Id: CAGfcS_m2QmqiA+ag-sJa3zitUGtCddmvJNGcQjCxxDSx7auxJw@mail.gmail.com
In Reply to: Re: [gentoo-dev] newsitem: initramfs required on Linux systems with separate /usr by "Michał Górny"
1 On Wed, Sep 25, 2013 at 4:35 AM, Michał Górny <mgorny@g.o> wrote:
2 >
3 >>
4 >> What about busybox[sep-usr]? Is that still supported or is everyone with
5 >> separate /usr forced to use an initramfs?
6 >
7 > I'd say it's supported as long as it gives a compatible end result.
8 > I suspect that the number of cases supported by that is less than those
9 > supported by a complete initramfs.
10 >
11 > However, I'd say the support is mostly the maintainer's discretion.
12 > As long as busybox maintainers want to support that, it should work.
13 > But don't expect Gentoo developers to check whether that work or
14 > encourage users to use that.
15 >
16 > I think we used to call that 'early boot mechanism' in the past, but I
17 > guess just 'initramfs' is easier for users.
18
19 If Gentoo actually offered some kind of formal support I'd be more
20 concerned about exactly what is and isn't supported, but for the most
21 part what is and isn't supported is a gray area that varies by
22 developer, with perhaps some hard boundaries at the extremes.
23
24 We had the conversation of whether mixing keywords was "supported" and
25 ended up basically where we seem to be with early boot mechanisms.
26
27 If somebody wanted to run with a separate /usr I would recommend they
28 use an initramfs. That doesn't mean that there aren't other ways of
29 solving the problem. However, an initramfs is what would end up in
30 the handbook/etc as it is probably the most straightforward solution.
31
32 I guess the questions is whether we really need to advertise the
33 alternatives. 95% of users are probably going to use an initramfs
34 anyway. If some prefer an alternative solution they're likely to
35 already be reading -dev and so on and probably are already using
36 busybox.
37
38 If the maintainers of the busybox-based solution want to plug their
39 option (and deal with the questions/issues when they arise) I have no
40 objection as long as it doesn't add much to the news item.
41
42 Rich