Controversial SOA values ?

Jim Reid jim at rfc1035.com
Wed Dec 29 11:40:46 UTC 1999


>>>>> "Michael" == Michael Milligan <milli at acmebw.com> writes:

    Michael> This just backs up my point.  If you can't get your (or
    Michael> a) master server back online within a week, you've got a
    Michael> bad or incomplete backup plan.  A good disaster recovery
    Michael> plan can handle all of these scenarios and have a master
    Michael> back online well within a week.

Indeed, but the difficulty is knowing that your recovery plan takes
account of all possible problems, some of which may be unforseen as
well as out of your control: like gales in the Atlantic stopping the
cable ship pulling up the broken line from the sea bed. BTW, the folk
who worried about a jumbo jet in the computer room had to recover from
that in an hour, and - another important point - they tested those
procedures regularly. They didn't use DNS, but that's another story.

I agree with you that a week should be long enough to get the master
server back. However I'd choose a longer expire interval - a month or
so - just to give more breathing space. After all it doesn't hurt. And
as Barry Margolin pointed out, there are ISPs who have clueless
customers that can't or won't fix things in a week. These customers
probably don't have recovery plans either. If you're the ISP, what do
you do? Try to keep the service going to these customers or leave them
to fester in their self-imposed mess?



More information about the bind-users mailing list