Re: Sites that block dynamic/dialups
From: D. Stussy (kd6lvw_at_bde-arc.ampr.org)
Date: 11/24/03
- Next message: Ruben: "Re: Can't get D-Link DFE-690TXD card to work w/ linux"
- Previous message: Menno Duursma: "Re: TFTP in Slackware-9.1 not working?"
- In reply to: Rob van der Putten: "Re: Sites that block dynamic/dialups"
- Next in thread: Rob van der Putten: "Re: Sites that block dynamic/dialups"
- Reply: Rob van der Putten: "Re: Sites that block dynamic/dialups"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Mon, 24 Nov 2003 13:35:10 GMT
On Sat, 22 Nov 2003, Rob van der Putten wrote:
> "D. Stussy" wrote:
> > ...And how does this work when a domain has an assymetrical mail handling system
> > - i.e. dedicated INBOUND and OUTBOUND services on different machines? An
> > outbound server would be connecting to you, but since it has no inbound mail
> > service (accessible from the Internet at large; just the internal dial-up
> > network), there's nothing to connect to. Some larger ISPs follow this model.
> > This could also be true if the message is being relayed - and the relay paths
> > are different for inbound and outbound mail handling.
> >
> > Now, if you meant to say that the verification connection would be to ANY server
> > pointed to by an MX record of that domain (or the host possessing the A record
> > if no MX records are present), that could still work in the assymetrical model.
> > However, relays that do not validate the username portion of the mailbox (but
> > merely "store and forward") will accept ANYTHING there, so I don't see how that
> > validates the mailbox. At best, it validates only that the domain is reachable.
>
> It uses the MX.
OK, but that isn't what you initially said.... You were going to use the IP of
the incoming connection. [Looks like you found you had to change that.]
> In my experience, systems that first accept the mail and then reject it
> are quite rare.
Being rare doesn't mean that you can ignore that case.
> You can detect such systems by testing a random address, such as a
> string based on epoch. The result of such tests can be cached.
Now you're increasing your overhead by doing a second test....
> > There may be other cases where this breaks also, such as autogenerated mail from
> > various non-mail servers - e.g. my ISP sends me an e-mail when someone signs my
> > web site guestbook. One can't generally e-mail back a service such as that.
>
> Such from addresses should have an alias. So should all system users.
And if those administrators (who need not be mail server admins) choose NOT to?
> > Why not fatal DNS errors ("NXDOMAIN") being fatal, and transient DNS errors
> > being transient?
>
> Sounds logical.
> I'll look into this.
>
> > And a reject isn't? All rejects of any relayed message cause bounces. You
> > don't necessarily know that a message has been relayed previously (depends on
> > when your system issues the reject - before or after the "DATA" subcommand, and
> > if after, how many "credible" Received: headers there have been). The only time
> > a reject doesn't cause a bounce is when the connection is between the e-mail
> > authoring client program and the [first] server. However, as the recipient
> > system, you don't control that. Granted, you issue a reject and some other
> > system not yours generates the bounce, but there's still a bounce. A spammer
> > isn't going to connect his client directly to your server, because that allows
> > you to read the IP address he's using at the time and therefore trace him.
>
> A sending MTA might impose restrictions on the the envelope from being
> used, in which case a false from would have to be in the same domain.
> It may also impose restrictions on the combination of origin host and
> destination address. In fact, without that, it acts as an open relay.
I don't see how that equates to an open relay. By operation, ISP's relays are
at a minimum CLOSED relays for their own customers, and it seems to me that your
above comments are true for EITHER type of relay.
- Next message: Ruben: "Re: Can't get D-Link DFE-690TXD card to work w/ linux"
- Previous message: Menno Duursma: "Re: TFTP in Slackware-9.1 not working?"
- In reply to: Rob van der Putten: "Re: Sites that block dynamic/dialups"
- Next in thread: Rob van der Putten: "Re: Sites that block dynamic/dialups"
- Reply: Rob van der Putten: "Re: Sites that block dynamic/dialups"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|