Re: Stupid relay question.

Keith wrote regarding 'Re: Stupid relay question.' on Thu, Oct 21 at 09:3=
9:
> On Thu, 21 Oct 2004 15:27:12 +0100
> Stefan Morrell <stef@mort.level5.net> wrote:
>=20
> > On Thu, Oct 21, 2004 at 04:21:33PM +0200, Magnus B=E4ck wrote:
> > > Say what? You can specify any lookup table in mynetworks, including
> > > hash (as long as your Postfix is built to support Berkeley DB). Rea=
d
> > > the documentation.
> >=20
> > This list, June 2003
> >=20
> > "To be precise: you can't use network/mask patterns in a hash table.
> > If you want to use a hash table for mynetworks, then you have to
> > enumerate individual IP addresses.
>=20
> If you are using a CIDR netblock you would either need to have each and
> every address in the netblock in the table or have some clever lookup
> that recognised 123.1.2.3 was in 123.1.0.0/16.=20
>=20
> The first is simply an automated expansion of what you would need to
> type in. =20
>=20
> I'll leave it to you to try and work out how to do the latter on a
> hashed table. I hope you have plenty of spare time if you accept the
> assignment.

given address =3D 123.1.2.3
first look up 123 (or 123.0.0.0/8, I suppose)
then look up 123.1
then look up 123.1.2
then look up 123.1.2.3

Granted, that will only work for /8 /16 /24 /32, but can't postfix alread=
y
do that for some lookups, or am I thinking of something else (apache,
perhaps)? One could always calculate all of the subnets that an IP exist=
s
in, and look up those, but that'd be quite a bit of overhead for the mini=
mal
gain. OTOH, just looking up the sub-IPs isn't much extra work, in an
absolute sense. I'd imagine my stupid regexp:/header_checks file is far
more significant to performance. :)

--Danny