Re: [cobalt-users] Reset serial Numbers in DNS to 1??

[flash22@xxxxxxx]

On Thu, 15 Feb 2001, Jeff Lasman wrote:

> flash22@xxxxxxx wrote:
> 
> > i Really don't think you want to do that -/
> 
> I can think of lots of good reasons to do it.  For example, making a
> mistake that makes the serial number too high, too long, or
> incomprehensible.

Hmm, it can't be 'too high' it's modulo, so it's supposed to wrap..
if it's 'too long' it should error out when the zone is loaded...

> 
> > The purpose of serial numbers is so other nameservers can check against
> > yours, setting them to very low numbers has interesting side effects, as
> > does making them smaller than they used to be...(remember you are gonna
> > increment them for new changes)
> 
> The purpose of serial numbers is so that slave nameservers will update
> properly.  That's all.

other/slave, yes..see below

> 
> > They should just have a number representing the date...plus a change
> > count...unless you've changed them a few hundred times in one day -/
> 
> While that's a good form of serial number, it's hardly required.  And
> lots of places don't do it that way (we do, however).

However he was refering to Raq...which as far as i know generates the
serial numbers in this format ...thus my confusion as to why he
was saying they had 'gotten out of hand' ..

Perhaps i should have said 'they should look like...'
> 
> > Why do you need them to be in sync with other nameservers? are you running
> > multiple masters instead of master/slave? nasty ;0
> 
> They're only for slaves.  Multiple master don't have to have
> synchronized serial numbers, though they should, especially if you
> populate them by copying the files (via ftp, for example); as many
> ISP-based DNS servers do it.

Tho in that case i would assume he wouldn't care about the serial
numbers...but i have learned not to assume knowledge on this list , thus
the trailing question just in case...:)

I would note an annoying ambuguity in the rfc's about serial's, they are
*required* to increment, and slaves are *required* to compare for exact
match, but the rfc's don't seem to say anything about what the slave is
supposed to do in response to getting a zone that's slightly older than
the one it already has...presumably it would be justified in ignoring it,
asking for it, or just plain complaining about it...thus my hesitation
about setting serial numbers down by small ammounts...

Stray strangness: rfc1892 declares that the use of 0 as a 'special' serial
number isn't allowed, noone told isc i guess ;)

There is also a interesting requirement that master nameservers are
required to refuse to load master zones with 'lower serial numbers' , tho
if you think carefully about this, it's not actually possible, unless you
are only reloading, how is the master going to know what the serial number
in the zone was the last tie it was running -/


gsh

DNS is simple, anyone can do it..it should be like a toaster...