memget errors in bindlog

[Kevin Darcy]

D. J. Bernstein wrote:

> SysAdmin writes:
> > It does not flush its own cache, (first in, first out) it just dies.
>
> Right. That's how BIND has always worked. According to RFC 1536, pre-TTL
> expiration is unfathomable evil, and you're supposed to buy more memory.
> ``Come on, memory is cheap!'' it says.

RFC 1536 is a 1993-dated Informational RFC, neither BIND nor any other
DNS server implementation is bound by its recommendations, nor, given its
age, is it particularly relevant to today's Internet IMO.

Having said that, however, I agree that crashing and burning is hardly a
reasonable way to deal with memory exhaustion. Can someone comment on
whether BIND 9 is/will be any better in this regard?


- Kevin