On Mon, Mar 10, 2003 at 09:48:41AM -0500, Rick Macdougall wrote: > Just wondering at the moment if we want to differenciate between User_prefs > (ie read) and Bayes DB (ie read/write) files, for example, the -x flag for > no user_prefs.
Yeah, definitely; they've got very different security implications, for example. Another issue, BTW, is if the spamc side may have info that the spamd side needs to know; for example, consider the "transmitting $HOME" case which has come up in the Bugzilla. Imagine this situation: - the MTA maintains a database, mapping recipient address to a home dir. This mapping is not easily computable, ie. maybe the home dir contains a random number or something ;) - the machine running spamd can access that home dir - spamd cannot access the recipient-to-home-dir database. (This btw is basically the case in a current open bug in the Bugzilla, I think Communigate is that MTA). In this case, the MTA *needs* to pass the info along to spamd, as part of the spamc connection. I suggest we allow this to be passed as part of the "username", which the user-info plugin (the aforementioned .pm file) can decode. In other words the sysadmin can set up spamc to be called as spamc -u [EMAIL PROTECTED]:/path/to/wierd/homedir ... < msg > filtered and the spamd user-info plugin decodes that -u string. --j. ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
