(Linux FC5, sendmail 8.13.7)

The use made by "check_mail" of the "From:" tag in the access database is
nicely flexible, allowing (for exmaple) entries both of simple domains
(e.g. "dom.ain") and full email addresses ("a.person@dom.ain"). For
instance the "dom.ain" style entries can be imagined as having the
high-level effect of pattern-matching anything at that domain.

(Hold that flexibility in mind and step sideways...)

For some time we've had our own "check_compat" ruleset, and it would seem
worthwhile replacing this with the supplied "FEATURE(compat_check)", if
reasonably possible.

But this feature seems only to allow exact names:
Compat:me@here<@>them@there

As I read it, it doesn't permit the sender or recipient to be some sort of
pattern (domain only, sendmail classes, etc). Our own rules make
extensive use of things such as "$j", "$=R" to determine OK/ERROR result,
including treating <@$+.$=R.> differently from <@$=R.>

Any thoughts? Pointers to descriptions and examples of doing such
pattern-like things via sendmail's own "FEATURE(compat_check)"?


--

: David Lee I.T. Service :
: Senior Systems Programmer Computer Centre :
: Durham University :
: http://www.dur.ac.uk/t.d.lee/ South Road :
: Durham DH1 3LE :
: Phone: +44 191 334 2752 U.K. :

David Lee wrote:
> (Linux FC5, sendmail 8.13.7)
>
> The use made by "check_mail" of the "From:" tag in the access database is
> nicely flexible, allowing (for exmaple) entries both of simple domains
> (e.g. "dom.ain") and full email addresses ("a.person@dom.ain"). For
> instance the "dom.ain" style entries can be imagined as having the
> high-level effect of pattern-matching anything at that domain.
>
> (Hold that flexibility in mind and step sideways...)
>
> For some time we've had our own "check_compat" ruleset, and it would seem
> worthwhile replacing this with the supplied "FEATURE(compat_check)", if
> reasonably possible.
>
> But this feature seems only to allow exact names:
> Compat:me@here<@>them@there
>
> As I read it, it doesn't permit the sender or recipient to be some sort of
> pattern (domain only, sendmail classes, etc). Our own rules make
> extensive use of things such as "$j", "$=R" to determine OK/ERROR result,
> including treating <@$+.$=R.> differently from <@$=R.>


Can you make your rulesets available as a patch and/or m4 file?

Also check_mail only has access to recipient information if you are
using delay_checks

On the other hand, check_rcpt always has access to sender information.

On Tue, 19 Sep 2006, jmaimon@ttec.com wrote:

> David Lee wrote:
> > (Linux FC5, sendmail 8.13.7)
> >
> > The use made by "check_mail" of the "From:" tag in the access database is
> > nicely flexible, allowing (for exmaple) entries both of simple domains
> > (e.g. "dom.ain") and full email addresses ("a.person@dom.ain"). For
> > instance the "dom.ain" style entries can be imagined as having the
> > high-level effect of pattern-matching anything at that domain.
> >
> > (Hold that flexibility in mind and step sideways...)
> >
> > For some time we've had our own "check_compat" ruleset, and it would seem
> > worthwhile replacing this with the supplied "FEATURE(compat_check)", if
> > reasonably possible.
> >
> > But this feature seems only to allow exact names:
> > Compat:me@here<@>them@there
> >
> > As I read it, it doesn't permit the sender or recipient to be some sort of
> > pattern (domain only, sendmail classes, etc). Our own rules make
> > extensive use of things such as "$j", "$=R" to determine OK/ERROR result,
> > including treating <@$+.$=R.> differently from <@$=R.>
>
>
> Can you make your rulesets available as a patch and/or m4 file?

Thanks for your reply, and my apologies for the delay in this
acknowledgement.

At present our ruleset "check_compat" ruleset is relatively simple, but is
both non-flexible and also has too much in-built knowledge of our domains.
So we cannot easily do routine updates based on users or domains, and
those rulesets are no good for use by other site admins.

I was hoping to have (or create) something more general, so that none of
the domain/user specific knowledge would be in the ruleset but instead all
be in the frequently updateable "access" (or similar) database(s).

At present (do I understand correctly?) "Compat:" can only be exact pairs:
Compat:me@here<@>them@there

where "me@here" and "them@there" are full, exact email names.

Taking that as the baseline, against which the comments below highlight
the variation, I was looking for the ability of:

(1) Compat:@here<@>them@there # (from) match: anyone "@here" (not just "me")
(2) Compat:@me@here<@> # (to) match: every-one @ every-there
(3) Compat:@$=m<@>them@there # (from) match anything in class "$=m"

etc.

I can probably code up rules for cases (1) and (2) and their obvious
extensions, probably modelled loosely on the rulesets that drive
"virtuser". But how to do something like (3)?


Does any of this already exist? Does anyone have something nearly there?

It seems a sound idea which we (the sendmail-using site-admins reading
this newsgroup, self included) ought to consider incorporating somehow.


--

: David Lee I.T. Service :
: Senior Systems Programmer Computer Centre :
: Durham University :
: http://www.dur.ac.uk/t.d.lee/ South Road :
: Durham DH1 3LE :
: Phone: +44 191 334 2752 U.K. :

On Thu, 28 Sep 2006, David Lee wrote:

> [...]
> At present (do I understand correctly?) "Compat:" can only be exact pairs:
> Compat:me@here<@>them@there
>
> where "me@here" and "them@there" are full, exact email names.
>
> Taking that as the baseline, against which the comments below highlight
> the variation, I was looking for the ability of:
>
> (1) Compat:@here<@>them@there # (from) match: anyone "@here" (not just "me")
> (2) Compat:@me@here<@> # (to) match: every-one @ every-there
> (3) Compat:@$=m<@>them@there # (from) match anything in class "$=m"
>
> [...]

I didn't proof-check that carefully enough. Try again:

(1) Compat:@here<@>them@there
# (from) match: anyone "@here" (not just "me")

(2) Compat:me@here<@>
# (to) match: every-one @ every-there

(3) Compat:@$=m<@>them@there
# (from) match any 'me' @ any local 'here' in class "$=m"

> [...]
> Does any of this already exist? Does anyone have something nearly there?
>
> It seems a sound idea which we (the sendmail-using site-admins reading
> this newsgroup, self included) ought to consider incorporating somehow.

--

: David Lee I.T. Service :
: Senior Systems Programmer Computer Centre :
: Durham University :
: http://www.dur.ac.uk/t.d.lee/ South Road :
: Durham DH1 3LE :
: Phone: +44 191 334 2752 U.K. :

David Lee wrote:
> On Thu, 28 Sep 2006, David Lee wrote:
>

> (3) Compat:@$=m<@>them@there
> # (from) match any 'me' @ any local 'here' in class "$=m"

In your ruleset match on $=m and update your lookup tag based on that.
When you build your db entires, use that tag.

You cant pass a class directly to a map for lookup.

On Fri, 29 Sep 2006, jmaimon@ttec.com wrote:

>
> David Lee wrote:
> > On Thu, 28 Sep 2006, David Lee wrote:
> >
>
> > (3) Compat:@$=m<@>them@there
> > # (from) match any 'me' @ any local 'here' in class "$=m"
>
> In your ruleset match on $=m and update your lookup tag based on that.
> When you build your db entires, use that tag.

Thanks. Do you mean match ruleset LHS on "$=m", and in the (successfully
matched) corresponding RHS put some sort of unique character string (e.g.
"<M_MATCH>"), and have the db constructed using that "<M_MATCH>" string?


> You cant pass a class directly to a map for lookup.

Thanks. (I was aware of that, but it never does any harm to mention the
obvious!)



--

: David Lee I.T. Service :
: Senior Systems Programmer Computer Centre :
: Durham University :
: http://www.dur.ac.uk/t.d.lee/ South Road :
: Durham DH1 3LE :
: Phone: +44 191 334 2752 U.K. :