[quagga-dev 11679] Re: link detect enable

David Lamparter equinox at diac24.net
Tue Oct 28 22:56:53 GMT 2014


On Tue, Oct 28, 2014 at 01:17:49PM -0500, eric.zuck at amacat.com wrote:
> On 2014-10-28 10:44, Paul Jakma wrote:
> > On Tue, 28 Oct 2014, Vincent JARDIN wrote:
> > 
> >> hacky, but could work, let's have a 
> >> getenv("ZEBRA_INTERFACE_LINKDETECTION") that can change the behavior.
> > 
> > Or just a have a global enable by default flag.
> > 
> > Thing is, then there's still the "why does our default hurt the common
> > case?" question.
> > 
> obligatory xkcd: http://xkcd.com/1172/
> > regards,
> 
> IMO, if you hide function behind environment settings, then you MUST 
> also dump the environment via the show commands. Otherwise, the config 
> _won't_ give an accurate representation of the behaviour (two identical 
> configurations with different behaviours....)

I fully agree, tbfh I think doing anything like this through the
environment is a horrible idea.

> What about punting the default value to auto-tools: allow the current 
> behaviour to be built via a ./configure --disable_auto_link_detect or 
> some such -- the suggested value is not good as it kind of suggests that 
> auto_link_detection is functionally disabled, and what is wanted is that 
> the default value is OFF..)

That ultimatively leads to the same situation:  a switch that isn't
visible in config, being set by the distribution's packagers...

The best thing I can think of is to add a command like:

	quagga defaults 0.99.23

which would go on top of the configuration and select the set of
defaults to be used;  if not present it'd be written with the initial
version used to create the config, so from then on you keep the defaults
of that version.

(It's not exactly bound to released versions, maybe we should use a date
code, or actually a separate version number.)

(We could already have used that for draft-idr-error-handling in BGP,
which is the last big behaviour-changing patch I remember)


If that wasn't obvious from the above suggestion, I'm strongly against
changing Quagga's behaviour between versions in ways that break
reasonable user setups.  There is of course a "too much" on this, as in
the xkcd comic above, but this isn't such a case in my opinion.


-David




More information about the Quagga-dev mailing list