TOSS_BOGUS_NEWSGROUPS

    KEYWORDS (like INN doesn't have enough things holding up the main 
              loop already....)

    Path header propagation

    apb mega patch, with worker limits, including RCHAN stuff.

    background renumbering

    ctlinnd lowmark

    CCflushfile

    Debugging and dmalloc hooks in inndstart.c

    Innd initial prority (inndstart.c)

    Priority setting in innd.c

    PerlMode() call when shutting down

    Much better usage diagnostics.  (Keep these!)

    Poison paths

    Matt Dillon's mmap active and NO_MMAP_ACTIVE (nnrpd/group.c)

    Optimization in article.c::OVERfind when last article number was 
        missing.

    Change to different DONT_LIKE_PULLERS code

    X-trace X-complaints to headers.

    ftp method in actsyncd.  (Seems useful.)

    Logging of "think it's a dotquad" message. (rc.c)

    Log level of shutdown in innd CleanupAndExit

    sigpipe-handler and no logging of "Connection reset by peer"
               (netscape wins this one.)

    A change in the order of the response to ARTICLE, HEAD, STAT.
        (I don't understand this.  I know it was probably wrong
         before, but changing something that is so embedded in
         the protocol seems likely to break things....  Can someone
         explain this one to me?)
        The code is in nnrpd/article.c::CMDfetch
#mibexp#ifexp XINN160B3  XINN1P6B2 INN160FE
         Reply("%d %ld %s %s\r\n", what->ReplyCode, art, av[1], what->Item);
#mibexp#elseexp
         Reply("%d %ld %s %s\r\n", what->ReplyCode, art, what->Item, av[1]);
#mibexp#endexp 460