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

Re: Bug#673688: confusing NEWS.Debian message



(Barging in from d-l-e as usual)

Gaudenz Steinlin wrote:
[...]
> I'm not sure if this change really warants an NEWS.Debian entry as for most
> people it does not change anything and as far as I understand it this is a 
> new additional feature, not a change in ifupdown behavior. Is it really 
> necessary that every Debian user sees this message?

(Or at any rate, every sysadmin responsible enough to use
apt-listchanges...)
 
> Below I have included an attempt to reformulate the message to make it easier
> to understand. But I'm not an english native speaker and might text might be
> far from optimal. Therefore I also CCed the debian-i10n-english list on this
> report.
> 
> Original message:
> ifupdown (0.7~rc1+experimental) experimental; urgency=low
> 
>     If ifup or ifquery is called with the --all option, if doesn't just
>     bring up all interfaces marked as "auto", but all interfaces of a
>     specified class, 'auto' by default. For the most uses, this doesn't
>     change anything, but lets all the interfaces of a specific class to be
>     brought up or queried.

There are a couple of tiny grammar glitches there, and a fiendishly
subtle distinction between "auto" and 'auto'.  More importantly, when
it says "ifup or ifquery", it seems in fact to mean "ifup, ifdown, or
(the new addition) ifquery".

Thanks for sneaking in /sbin/ifquery, by the way!

>     If ifupdown is called with the --all option, before or after doing

I spent a while trying to work out why it should be drawing this
contrast between "If ifup or ifquery is called with the --all option"
and "If ifupdown is called with the --all option".  But I don't think
that's intended - when it says "ifupdown" here it doesn't mean (say)
/etc/init.d/ifupdown, it's using the name as a shorthand for "the
single binary hardlinked as ifup and ifdown".  That seems like an
unnecessarily obscure implementation detail, especially when in fact
ifquery now provides a third name for the same file!

>     anything to the interfaces, it calls all the hook scripts (pre-up or
>     down) with IFACE set to "--all", LOGICAL set to the current class
>     specified by the --allow option (or "auto" if it's not set),
>     ADDRFAM="meta" and METHOD="none".
>
>  -- Andrew O. Shadura <bugzilla@tut.by>  Tue, 17 Apr 2012 01:05:42 +0200

I suppose I'd make that
      ADDRFAM set to "meta", and METHOD set to "none".
But I'd also prefer to reshuffle to put the complicated one last.

It's unclear how this is a change from the previous situation, but I
notice a similar paragraph has been added to interfaces(5) (without
changing the "5 April 2004" footer, by the way).  Wouldn't it make
more sense to say something along the lines of "See interfaces(5) for
details"?
 
> My reformulation attempt:
> ------
> The --all option to ifup and ifquery can be combined with the --allow 
> option to bring up or query all interfaces of a specific class. The 
> default class if no --allow option is present is "auto". Therefore
> for most use cases this does not change anything.

It might be better if we run those sentences together:

  The --all option to ifup/ifdown/ifquery can now be combined with the
  --allow option to address all interfaces of a specific class. If no
  --allow option is present, the default class is "auto", so for most
  use cases this does not change anything.

> If ifupdown is called with the --all option, before or after doing
> anything to the interfaces, it calls all the hook scripts (pre-up or
> down) with IFACE set to "--all", LOGICAL set to the current class
> specified by the --allow option (or "auto" if it's not set),
> ADDRFAM="meta" and METHOD="none".

At least make it clear what commands it's talking about.  I'm putting
ifdown in parentheses throughout with an implied sprinkling of
"...respectively".

  When ifup (or ifdown) is invoked with the --all option, before (or
  after) doing anything to the interfaces, it calls all the pre-up (or
  down) hook scripts with IFACE set to "--all", ADDRFAM set to "meta",
  METHOD set to "none", and LOGICAL set to the current class specified
  by the --allow option (or "auto" if it's not set).

But as I say above, it's unclear how much of this is NEWS.

Or here's an ultra-squeezed version of the whole thing:

  The --all option to if{up,down,query} can now be combined with the
  --allow option to act on all interfaces of a specific class (still
  defaulting to the class "auto"). See interfaces(5) for details of
  how hook scripts are called.
 
-- 
JBR	with qualifications in linguistics, experience as a Debian
	sysadmin, and probably no clue about this particular package


Reply to: