I have a bunch of ditto's to add to Ned. Our implementation also has
global scripts, stored in LDAP, that are run very similarly to what
Jutta's draft specifies.
Likewise our product currently permits 5 different styles of sieve script. One
owned by the server administrator run before the message data is accepted
through SMTP, one when the message reaches the domain, and the other 3 before
the message data is written to the end users mailbox.
Some actions operate differently depending on which script they are used in,
for example fileinto in the server script files to the file system, not to a
mail folder.
We found we needed a set of security related rules that the user could not opt
out of, and a set of policy rules that they could, so the domain administrator
controls a sieve script run immediately before, and one immediately after the
end users' own script. draft-degener-sieve-multiscript-00 documents a useful
discussion of the complexities between "stop processing ANY more scripts" as
opposed to "stop processing this script" which is certainly useful and might be
worth keeping.
Nigel