On Wed, Feb 20, 2002 at 10:34:20AM -0500, Charlie Summers wrote:
That is, gc-d is _supposed_ to send the message back to gc, since gc should check for a unique Message-ID: header field, and if it finds a duplicate, should NOT send out the mail a second time. (See rc.submit; search for "Eliminate duplicate submissions" to review the recipe.) But if it _doesn't,_ it should send out the mail (since the poster sent it to the digest list).
We use an ancient version (3.10) at eskimo.com, so forgive me if this has all changed in more recent versions, but on our setup: The digest accepts posts ONLY from the list (and the maintainer, so I can forward things that confuse my quotation-removal recipes, which operate in digest/rc.local.s10). That is, the digest/accept file has just those two entries. So presumably, the digest can assume that the list has already done duplicate screening. The digest has its own dist file which is linked to list/accept2, so digest members can post, but they must post to the list. This makes sense to me, as the digest is comprised of list articles. Has this strategy changed in more recent versions? I've often thought it'd be nice if SL could be made clever enough to accept posts to the digest from legitimate subscribers, just silently passing them to the list, on the theory that the poster wasn't clever enough to realize the digest is a separate entity. Most of my subscribers eventually figure it out, but for some it's pretty painful... :) It's not something I've felt strongly enough to impliment myself. If it's in a newer SL version, I might lobby my ISP to install it. Jim