http://bugzilla.spamassassin.org/show_bug.cgi?id=3661





------- Additional Comments From [EMAIL PROTECTED]  2005-03-15 02:05 -------
Subject: Re:  Request for HTML de-obfuscation of invisible SPAN's

> however I see Loren's point that adding a new qualifier (like the
"something" in
> "header FOO NameOfHeader:something") is a bad idea, too, from a
rule-developer's
> usability perspective.  each qualifier is _really_ a new rule type, just
disguised.

Even worse would be "body FOO blah" followed by a TFLAGS to change what the
test was really looking at...

> in that case, adding new rule types can be a good idea -- if we can avoid
the
> overhead of new rule type hashes to track them.

In most of the cases I can currently think of where I'd like to see new test
types, the test types are almost always body variants of some sort.  The
'cleanbody' I mentioned that started this discussion, or 'body:full' that
really has the whole text body in one string, or something similar.

All of these are essentially built from existing pieces, either by simple
joining or kinda simple filtering.

I think I'd be inclined to have a small hash that simply gets the test types
found in rules contributed to it as they are found.  Then at the top of the
evaluation loop in PMS, I'd look in the hash or boolean flags array or
whaever method was chosen, and generate the requested oddball test items on
the fly before running all the tests.  The tests can still be grouped by
header tests and body test etc as desired, and the necessary data will be
available.

The only other hard part, it seems to me, would be having a
hash/group/regex/whatever in the rule scanning/parsing code that can
recognize the various test types and classify them as header or body test
type of thing, and flag each type of test item found for later use during
evaluation.





------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to