[bind10-dev] config experiment

[Francis Dupont]

Another point about config: we had hard discussions about the style of
running config in AFTR/CGN. The choice was between the Cisco style with
startup-config and running-config where the CLI modifies the
running-config (*), and the anti-Cisco style where all changes are done
in the config file and committed through a reload of the whole config (**).

As far as I can see the style in BIND 9 is anti-cisco but I don't know
if you have a clear choice for BIND 10.

Francis_Dupont at isc.org

PS:
(*) the Cisco style is known to lead to situations where the idea about
the running-config is different to the real one. So scripts are
developed to track these differences, not very easy because of the loose
syntax of Cisco config file and still relying on an exact output by
the box of its current running-config. It is known to be a possible
real mess so some adopted (or enforced :-) the opposite style.

(**) the anti-Cisco style is hard to implement without stopping the
service (I believe this is why Juniper which proposes it developed
fast-restart version of all routing protocols :-) and for some
critical parameters don't make sense without rebooting the whole
module (we have already some examples in BIND 9, just look at
parameter changes in caching servers where flushing the cache
after a reload :-).