[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#684726: ITP: check_v46 -- Icinga / Nagios plugin for dual stacked (IPv4 / IPv6) hosts



On Wed, 2012-08-29 at 14:35 -0700, Don Armstrong wrote:
> See Policy §7.2. If the depended-on package is not "required for the
> depending package to provide a significant amount of functionality"
> but is a package which "would be found together with the [depending
> package] in all but unusual installations", it should be a
> Recommends:, not a Depends:.
Yeah,.. but this is really very vague, isn't it? And AFAICS the
consensus on most distributions and also most Debian packages was that
depends is the relation that indicates that the package is required for
core functionality.


Therefor, and this is also in reply to Timo Juhani Lindfors mail before,
I personally would say roughly the following:
A depends on B if either of the following is true:
- without B, A cannot be installed/configured
- without B, A fails not gracefully  (i.e. corrupts data, segfaults,
doesn't work at all without warnings/hints)
- without B, using A wouldn't make sense at all (as it looses it's core
functionality), even if it would fail gracefully


I largely agree with what Russ, wrote below, I like to install only what
I really want, and giving information in the package description about
the Recommends and Suggests is a good way to help understanding.




But apart from all that,... my intention was no to ask for
"clarification" of the policy in that matters (even though that wouldn't
harm perhaps ;) )


A package like nagios-plugins-contrib is obviously not a meta-package.
It's however not a normal package either, as it contains different
programs from different upstreams, whose only "connection" is, that all
of them are nagios/icinga checks.
I call this "compilation packages".

IMHO the following is likely true:
- For some compilation packages it would be better if they were split
up, especially when one considers that Debian would ultimately package
"all" (or at least very much) of the available Nagios/Icinga pluings.
The compilation package would just get enormous big.

- Ask yourself, if you were the maintainer of these plugins and if you'd
decided to package them separately, would you have used Depends or
Recommends in many of the cases (I talk about e.g. the packages which
provide programs absolutely needed by the plugins)?
So I'd conclude, that Recommends was chosen in some cases only, to keep
the dependencies small (because people typically tend to not use each
program of a compilation package anyway),... and therefore this sounds
like "wrong" packaging[0] to me.


- Those who argue that this is a problem because one gets so many very
small packages,... we do this already,... with the font-* packages, for
example; and it seems to work.



Cheers,
Chris.

> 0: I wrote the above, not Bernd. Please try to avoid incorrect
> attribution when you quote e-mail.
Yeah sorry,... was a mistake when cutting out Bernd's comments.


[0] Of course this is not meant specifically "againt"
"nagios-plugins-contrib" and it's maintainers; so don't feel offended.
It's a general issue IMHO in current packaging practise :)

Attachment: smime.p7s
Description: S/MIME cryptographic signature


Reply to: