On Thu, Apr 25, 2002 at 11:45:33PM +0200, GOMBAS Gabor wrote:
> On Thu, Apr 25, 2002 at 01:39:07PM -0500, Scott Lamb wrote:
>
> > I started by making a patch to Postfix that would just add the
> > "X-RBL-Warning" header as some other MTAs do.
>
> IMHO think this is the Right Thing(TM).
IMNSHO this is a kludge. I did it because it's a minimalist solution my ISP
would be likely to apply for me to take advantage of. Now I'm looking to
solve the problem a better way, with something novice users could use.
> > It's more likely to get into graphical Sieve generators this way.
>
> I'd never use a Sieve generator that does not allow specifying a filter
> based on a custom header...
That's not what I'm talking about. You can do that now. The point of a
graphical editor is to make it easy. I'm thinking of a test dialog that has
options for common anti-spam lists. I.e., the categories I've listed here as
IP addresses as checkboxes with text descriptions and "What's this?" text.
That's a far cry from you telling it to check "X-RBL-Warning" for
/^relays\.osirusoft\.com \(127\.0\.0\.[2456789]\)$/.
> > (Would you want a wizard that would grep the contents of an experimental
> > header some MTAs use?
>
> It's a chicken-and-egg problem. If nobody uses it because it is experimental
> then it will remain experimental forever. If it becomes widespread it has a
> good chance to become standard.
It has no chance:
- People don't always run the latest versions of their MTAs. They often
don't configure them correctly. If they did, we wouldn't have any open
relays to block, and I wouldn't be posting this at all. For these reasons,
X-RBL-Warning will never be used by 100% of sites. And since there's no
announcement to tell if it's used by a given site or not, anything less
than 100% is not good enough.
- The X-RBL-Warning headers being generated are different. That's what
happens with experimental things. It would be a pain to hunt them down and
write something that would parse them all. If someone did, there would be
another type by the time they were done.
For a _good_, _reliable_ graphical wizard to exist, it MUST be possible to:
- tell when the dialog pops up if this is supported or not. Otherwise,
people will be confused by it when it doesn't work and there are poor/no
diagnostics. Possible with Sieve feature announcements. Not possible with
headers in the file, especially ones that only appear on spam.
- specify what RBL(s) to use, or at least know what ones are available.
- reliably obtain the information. Again, the different X-RBL-Warning
headers being generated would be a problem.
All of this is possible with the extension I'm attempting to create.
Configuration: the mail admin would set a variable ("rbl_received_header",
say) indicating which "Received:" header to parse, as Bob Finch suggested.
The "x-rbl" feature announcement is made iff this header exists. Its
presence indicates the admin has ensured that "Received" header contains the
correct address to run through the RBL and is in a format my parser can
parse. So the feature announcement is only made if what I've said above is
true. It could be used as the foundation for a wizard.
--
Scott Lamb
