Change of IP - Update time

[Steven Job]

> 2) Your end-user is (probably unbeknownst to him) cacheing your DNS
> either in his client program (browser) or operating system.

And this usually expires when the TTL expires.  An operating system that
does not do this is flawed.

 
> 3) Your end-user is on a network that's cacheing (again, probably
> unbeknownst to him) your DNS.

If nameservers are cacheing information then that is flawed also and it is
breaking the internet.  That's the whole point of TTL.  Why would that ever 
be ignored.  If there are nameservers out there that ignore TTL then I would
like to know about them.  The are delivering poison to the internet.

 
> 4) your end-user is getting DNS from a nameserver that caches DNS for
> overly long time periods ignoring your TTL.

My same point as above.


These are just my opinions though.


On Sun, 5 Jan 2003, Jeff Lasman wrote:

> 
> Peter Pfannenschmid wrote:
> 
> > I have the following problem: I would like to use a name server (which
> > is running fine and without problems) for some sort of dynamic DNS.
> > Let's assume we have the host test.mydomain.com with IP address
> > a.b.c.50; if I change this to a.b.c.51 and reconfigure the DNS server,
> > there may be a delay up to two days until some client at the other end
> > of the world gets the changed IP address when querying the host name.
> 
> There are several reasons for the delay:
> 
> 1) Your TTL for the record in question is overly long.
> 
> 2) Your end-user is (probably unbeknownst to him) cacheing your DNS
> either in his client program (browser) or operating system.
> 
> 3) Your end-user is on a network that's cacheing (again, probably
> unbeknownst to him) your DNS.
> 
> 4) your end-user is getting DNS from a nameserver that caches DNS for
> overly long time periods ignoring your TTL.
> 
> Of these, you can only control #1.
> 
> > I suspect that the reason for this is the fact that multiple name
> > servers can be involved in the resolution of a client's query, and
> > every one of them could have the old IP address cached. So it will
> > take some time until the new IP address is propagated. Please correct
> > me if I am wrong here.
> 
> See my notes above.
> 
> > On the other hand, there are services like dyndns.org which are
> > specialized in this kind of stuff; these guys manage to propagate a
> > changed IP address (a changed DNS record) throughout the world within
> > five minutes after the changes have happened on their DNS servers. My
> > question is: How do they achieve this?
> 
> They, like you, can only control #1.
> 
> > I am planning to experiment with the TTL and other parameters of my
> > zone file. But before I do so, I would like to hear the opinions of
> > the experts, since a too low TTL time may be considered as bad
> > practice by the community.
> 
> If you want to serve dynamic DNS you have no option than to use a low
> TTL, perhaps only seconds.  As you can see from reading my points above,
> that won't help in all situations.
> 
> Dynamic DNS is a tradeoff at best, and only works even reasonably well
> for most people when the dynamic IP# is changed only occasionally.
> 
> > So what? Is there any method to achieve my goal, or should I set the
> > TTL parameter to five minutes?
> 
> Perhaps even less.
> 
> > If the latter is the case, is there a
> > way for changing the TTL of the record for only one certain host of a
> > domain (other than making a new zone)? I am using bind 9.1.3.
> 
> Yes... read the fine manual <smile>.
> 
> Here's a hint, using my own domain name and a fictitious host:
> 
> fictitious.nobaloney.net.	1M IN A		63.108.93.101
> 
> The above example would give fictitious.nobaloney.net a ttl of one
> minute.
> 
> Jeff
>