Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!psuvax1!news From: flee@cs.psu.edu (Felix Lee) Newsgroups: news.software.b Subject: Re: What I'd like in a news expire program Message-ID: Date: 12 Mar 91 04:59:55 GMT References: <1991Mar11.201850.14997@mthvax.cs.miami.edu> Sender: news@cs.psu.edu (Usenet) Organization: Penn State Computer Science Lines: 28 Nntp-Posting-Host: guardian.cs.psu.edu I've been thinking about steady-state news, and I think all you need are low- and high-water marks on disk space usage. You provide disk space boundaries for various sets of newsgroups, like so: keep [0, 20] megabytes talk.all + soc.all + all.flame keep [50, 100] megabytes all which means that all of news should take no more than 100M, and noise groups should take no more than 20M of that. To deal with explicit expire date and exceptionally large articles, you can create subsets with set intersection: keep [0, 15] megabytes comp.all * (expires > 30 days) keep [0, 20] megabytes all * (size > 1 megabyte) I don't know if I want time boundaries. As long as news doesn't exceed space boundaries, I don't see a reason to expire it. Well, groups like misc.forsale might want a time boundary keep [0, 7] days all.forsale to keep the contents timely. Time boundaries might conflict with space boundaries. To keep things sane, I think all lower boundaries should be soft. If you say keep [3, ] days all it won't actually guarantee that news is kept 3 days, but it will ring an alarm if news does fall below 3 days in any newsgroup. Comments? Any flaws in this scheme? Any useful features missing? -- Felix Lee flee@cs.psu.edu