furuta (01/16/83)
Thanks to those who pointed out to me that net.periph was a misspelling. The editorial comments in personal messages to me could have been omitted. I've been thinking a little about the difficulties which the recent spate or rmgroup messages have caused for some net news administrators and readers. One particular class of problems seems to be caused when a remotely generated rmgroup message causes information to be lost. Another is that once a bad group name is introduced onto the net, subsequent postings tend to perpetuate the difficulty. Since the administrator's reaction to the first kind of problem is to prohibit remote rmgroups, the second kind of problem doesn't get corrected (ignoring that some people are running older versions of news, etc.). I'd like to propose a compatible change which I think will help remove some of those difficulties in the future. Specifically, change the semantics of the rmgroup message. Instead of physically removing the group, the rmgroup message should only prohibit local posting to the group. A subsequent newgroup message for the group would permit posting again. I suppose it should be an option (for each site to decide) as to whether or not articles coming in from elsewhere directed to the dead group were accepted or not. Personally, I'd prefer to receive these articles. Actual removal of the group should be accomplished by an expire-like daemon which would be given permission to delete groups which had been inactive for some period of time (say, 3 months--again, an option). Implementation could be accomplished by having the rmgroup deposit the name of the group into a file, say /usr/lib/news/deadgroups, checking that the group wasn't already in the file first, of course. inews would check this file before allowing posting of an article. Similarly, response to a newgroup message would involve checking this file. I think I prefer to have this file separate from the already existing ngfile as that provides a separation between the manually generated ngfile and the automatically generated deadgroups. Compatibility is retained since sites running older versions of net news would continue to react as always to rmgroup and newgroup messages. With these changes in place, I'd feel quite happy about undefining MANUALLY again. --Rick ...decvax!microsoft!uw-beaver!uw-june!furuta (uucp) ...ucbvax!lbl-csam!uw-beaver!uw-june!furuta or Furuta@Washington (ARPAnet)