Re: init.d scripts and LSB
* Craig Small <csmall@eye-net.com.au> [020505 20:19]:
> I have got bug #138251 which talks about the init.d script and how it
> is missing some nices things etc.
>
> Should Debian scripts be following the LSB and if so, why doesn't the
> policy either mention the LSB or have the same standards?
>
> This is more important for me as the dh-make maintainer as a lot of
> people use it to make their own packages so I like to get it right.
>
> I'm not on debian-policy, so please CC me.
I posted to the lsb-spec mail list and got some replies that are on
target. I attached them below. Check the archives of those lists to
see the full threads, but here are some key responses from Alan Cox and
Ted Ts'o. Chris Lawrence did a better job doing a complete and accurate
summary of the issue at hand than I did.
So in short, I was wrong. Technically the init script LSB
specifications are only applicable to lsb applications/packages.
However the init script names must be coordinated as Ted mentions.
I don't fully understand how this should/will function in practice.
My thinking was from a user/administrator perspective such that all init
scripts SHOULD have the same consistent set of functions. I was
thinking that if they are defined clearly in the LSB we might as well
use them. Yet this is a separate policy discussion that's beyond the
scope of the LSB and goes to my personal opinions about possible future
Debian policy.
Cheers,
--
-- Grant Bowman <grantbow@grantbow.com>
On May 07, Grant Bowman wrote:
> There's a discussion going on right now over on the
> debian-policy@lists.debian.org mail list. I am looking for validation
> in the specification itself that the LSB applies to systems as a whole
> and not to only *.lsb packages. This seems like a crazy premise to me,
> but I'm having trouble finding justifications in the specification to
> clearly demonstrate otherwise. Any suggestions?
* Alan Cox <alan@lxorguk.ukuu.org.uk> [020507 15:35]:
> Bit of both
>
> It applies to LSB compliant applications
> It applies to the OS in the sense that it must provide LSB defined features
>
> There has never been any intent that it should apply beyond that
* Chris Lawrence <quango@watervalley.net> [020507 13:49]:
> Just to be clear, the debate is over whether distribution-provided
> init scripts must comply with the Init Scripts section of the
> specification. My and others' position is that there is no such
> requirement, since LSB-conformant applications (as of LSB 1.1) cannot
> depend on the presence or absence of any particular init script on the
> system, so (a) they shouldn't care what command line arguments they
> accept and (b) they shouldn't care what exit codes they return.
>
> Now, I don't think there is any disagreement that init scripts
> provided by LSB-conformant applications must comply with the spec, as
> the spec provides an interoperable subset of capabilities for init
> scripts to use and an interoperable superset of command line arguments
> for init scripts to function normally.
>
> (Having said that, I wouldn't be opposed to a requirement in the
> future, if the interfaces were clearly defined and there was some good
> reason for LSB applications to be mucking with init scripts provided
> by the system... for now, though, that seems rather unlikely.)
* Theodore (Ted) Ts'o <tytso@mit.edu> [020507 15:07]:
> Yes, that's correct. The init scripts section of the distribution is
> only applicable for LSB applications.
>
> In particular, the commenting conventions were designed only for LSB
> applications. Attempts to generalize the Required-Start: headers,
> etc. to work for all init scripts, including system header files, is
> possible, but it makes the install_initd script far more complicated.
>
> The only thing which applies to system init scripts as well as the
> naming convention. Here, distributions should register all init.d
> script names that they use with LANANA, so that LSB-compliant
> applications don't choose init.d script names which conflict with
> distribution-provided init scrips.
--
To UNSUBSCRIBE, email to debian-policy-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: