$GENERATE or $EXEC?

[Paul Vixie]

"Cricket Liu" <cricket at acmebw.com> writes:

> That's an interesting idea, Andre.  And with $EXEC, you could replace
> $GENERATE with a small program and the appropriate right-hand side:
> 
> $EXEC /usr/sbin/named/generate 1-63 $ IN CNAME
> $.0-63.0.168.192.in-addr.arpa.
> 
> I guess someone would have to publish a new RFC extending the current
master
> file format to include $EXEC, too.

sorry to rain on the parade, but i'm going to resist putting this feature in
for the same reason i ended up resisting putting in read-zone-from-pipe (as an
alternate to "file ..." in "master" zones): there is no way to know from within
the server that reading the zone over again could have a different result. in
the pure "file"/$INCLUDE model there is a way -- the stat() posix call -- to
know that the zone ought to be reread.  if we use pipes, either at the config
file level or zone ($EXEC) level then we need a separate program that just
gives back the current SOA, or mtime, or whatever.  or we need standardized
arguments to the program so if it's called with "-justsoa" it has different
output.  this whole area is ripe for research but nowhere near development,
even advanced development.  (note that $INCLUDE currently isn't graphed and
we automatically assume that the zone could have changed if $INCLUDE is used;
this is however an implementation fault rather than a design fault.)

so at the moment i have to suggest that $EXEC be done in a perl or m4 pipeline
outside of the server.  $GENERATE is at least static with regard to stat()'s
reported mtime.
-- 
Paul Vixie <vixie at mibh.net>

"I sorta lean towards different both." --asp