Re: [YANG] mandatory & default

Phil Shafer wrote:
> "Randy Presuhn" writes:
>> One of the reasons for archiving configurations
>> is so that if a box fails, a replacement can be quickly made
>> to function in its place.
> 
> Loading the saved file with defaults won't make this
> any saner, safer, or stronger.  It increases the fragility,
> but incorporating into the saved file tons of stuff about
> which you don't care.  It something fails on these values
> that you don't care about, you are broken.
> 

this is not correct.
If you load a config with the full set of values, then
agent is forced to accept the value or issue an error.
This is not fragile.  This is debugging.

Fragile is when the new software image sees that no value
is given for the foo-knob, so it assigns a *different*
default value than the last SW image, and does not complain
at all.

Andy

>> If there is even a remote chance
>> that a default value might change between releases or on
>> different versions of a hardware platform, that value needs
>> to be included in an archive of the configuration.
> 
> The chance of a change in defaults (which annoy customers to
> no end) is minor compared to the change of adding features (which
> customers insist on) and the need to support downgrading (which
> customers insist on).
> 
> Consider this example from JUNOS-land.  We have a set of foo-options
> containers that are used to hold media-specific options under
> the [interfaces] statement.  These options are numerous, verbose,
> and rarely used.  If I add a new statement in release 9.0 and
> start emitting a boring default value for it, I can no longer
> load that config in any release predating 9.0.
> 
> In avoiding a rare bug (changing defaults), you'll break real
> operational usage.
> 
> Thanks,
>  Phil
> 
> 
> _______________________________________________
> YANG mailing list
> YANG@ietf.org
> https://www1.ietf.org/mailman/listinfo/yang
> 
> 

_______________________________________________
YANG mailing list
YANG@ietf.org
https://www1.ietf.org/mailman/listinfo/yang