ccTLD not using BIND

Simon Waters Simon at wretched.demon.co.uk
Tue Sep 3 07:06:45 UTC 2002


Sebastian Castro wrote:
> 
> Moreover, they have affirmed that BIND fails to answer
> between 2 to 5 percent of queries. I haven't found anything to confirm
> or deny this proposition.

I think any software will fail when resource starved, but I
haven't seen anything to suggest properly configured and
resourced BIND fails to answer queries. BIND is a tad resource
hungry, mainly due to attempting to load all the data into
memory at start-up, there is an arguable trade off between
performance and memory.

Similarly out of the box DJBDNS products may want recompiling to
increase some of it's default limits under particularly
exceptional loads, typical BIND covers the same thing with
changes to named.conf.

I have seen some non-BIND, non-DJBDNS, name servers give
responses that floored "dig", my guess is if dig didn't handle
them, most name servers would be perplexed, as for all it's sins
the ISC code is largely standard conforming.

> So, I came here to the Olympus of DNS knowledge, to find light among the
> darkness :)
> 
> I hope anyone can help or give me some advice (or references).

Technically the wrong newsgroup I think,
comp.protocols.tcpip.domains is the source of more alternative
DNS information.

I wouldn't assume anything done by ccTLD admins is "right",
without checking the standards, some of these groups are
surprisingly poor at DNS management, and not necessarily the
countries who might have trouble resourcing their DNS.

Choice of DNS server wouldn't be my main concern with running a
ccTLD, you have two products you can be sure will do the job,
now how are you going to manage the database of zones and
changes. Some of the fall out from that may dictate what name
server you choose, i.e. database back end, use of IXFR etc etc.


More information about the bind-users mailing list