[Vmail-discuss] Finally... vmail-0.4

Paul Makepeace Paul.Makepeace@xxxxxxxxxxxxxxxxxxx
Tue, 31 Jul 2001 01:39:45 -0700


On Tue, Jul 31, 2001 at 09:04:43AM +0100, Chris Lightfoot wrote:
> > 1. I noticed some SQL things,
> > 
> > In README,
> > 
> >   query = "select remote_name from forwarder, domain where local_part = '$local_part' and
> > forwarder.domain_name = domain.domain_name and domain.domain_name = '$domain'"
> > 
> > is equivalent to:
> >         select remote_name from forwarder
> >           where local_part = '$local_part' and domain_name = '$domain'
> > 
> > Ditto the other query.
> 
> True for Oracle, where presumably you have foreign-key
> constraints; but not for MySQL, where one could put rows
> into the forwarder table which didn't correspond to `real'
> domains. So from your PoV, yes; from ours, not quite,
> because of paranoia about audit trails.

I don't quite understand the point here or what you're trying to
safeguard. If it's malicious/accident-prone users adding dud entries to
forwarder then they could equally add to domain and I don't see what
good this does. Or am I missing something here?

That dual join is an extra expense. Which I haven't benchmarked... (this
whole email is mostly aesthetic concerns anyway).

> Well, I don't think that s/^_default_$/*/ is going to
> impose a big burden on your dump-to-CDB argument.

Yeah, yeah, I know :-) Just looked nicer.

> (I note in passing[1] that [RFC822] `*' is just as valid a
> local-part as `_default_'.

Some addresses are more valid than others, at least as far as RFC822
and the myriad interpretations of it by clients & servers are
concerned :-)

Paul <*@paulm.com>

-- 
It is simply a matter of work