Xref: utzoo news.admin:12411 news.software.b:6951 Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!sdd.hp.com!wuarchive!mit-eddie!bloom-beacon!eru!hagbard!sunic!mcsun!ukc!slxsys!ibmpcug!mantis!mathew From: mathew@mantis.co.uk (mathew) Newsgroups: news.admin,news.software.b Subject: Re: What to do about humongous postings Message-ID: Date: 22 Feb 91 17:16:53 GMT References: Organization: Mantis Consultants, Cambridge. UK. Lines: 66 mcb@reason.ig.com (Michael C. Berch) writes: > In the referenced article, mathew writes: > > Recently there have been a couple of instances of people posting enormous > > uuencoded files in one piece, causing various systems to crash. [...] > > People seem to think that something should be done. I suggest that any > > posting longer than a certain threshold length should be dropped, and an > > error report mailed to the originator [...] > > I think the statements about "getting around the limit" in the above > posting shows that the author seems to be confusing two "problems" [I > put "problems" in quotes as I am not willing to admit that either is > sufficiently obnoxious so as to require a software-based correction]: > > 1) Certain large individual articles may harm particular systems > by filling up disk partitions allocated to news or causing > transport-level (e.g., UUCP) problems; and > > 2) Certain individual posters are thought to abuse the network by > posting too many bytes (in the aggregate) in too short a period. No, I'm quite aware that these are two separate problems. I believe that there are a sufficient number of sites which have problems with large articles that large articles should be discouraged. You _could_ drop large articles only when they hit UUCP, but that would be extremely confusing to the user, not to mention annoying for those in the UUCP world who _want_ to get the articles, but need them split up. (I don't think software is up to intelligently splitting an article into multiple pieces depending upon content...) > Sites which are > affected by large articles should either install filters to protect > themselves, or drop those groups in which large articles often appear, > i.e., the binary and image groups. (1) It's not much use our dropping these articles _AFTER_ our system has spent half an hour downloading them from our feed site. I want inappropriately-long articles to be junked _BEFORE_ I pay for them and before they cause our system to crash. (2) We do drop the binary and image groups. I'm talking about incidents where people post entire movie scripts, core dumps or whatever to a discussion group. (3) Spool space is limited. UUCICO doesn't know how big a file is until it's too late. So to clarify what I am proposing: Single messages longer than a certain length (to be decided) which appear in any newsgroups should be dropped, and mail sent to the originator. Personally I can't think of any reason for single messages of more than (say) 128K in a discussion group, or even in a binaries group for that matter. I'd probably support a limit of about 32K for discussion groups. I believe that most incidents of this type are caused by people who do not understand that lots of us pay money to shunt Usenet articles around. This was certainly the case in comp.sys.acorn, where I received mail from people saying "if you don't like huge binary postings in a discussion group, just hit 'n' and read the next article". It was also the case in alt.fan.monty-python, where the naive user concerned was apparently unaware that posting 700K in one message could cause any problems. mathew.