Article View: gnu.emacs.vm.bug
Article #9179Re: Babyl
From: Calvin Clark
Date: Wed, 11 Mar 1992 01:28
Date: Wed, 11 Mar 1992 01:28
53 lines
3347 bytes
3347 bytes
In article <DJM.92Mar9194708@frob.eng.umd.edu> djm@eng.umd.edu (David J. MacKenzie) writes: > Processing MH folders a lot of overhead from opening lots of files > and wastes a lot of disk space because of minimum fragment sizes. > Other than those problems, they're ok. I agree with the statement about slowness due to finding and opening many files. The difference in disk usage between storing messages individually and storing them in one big file is usually not that significant, except for users who get a lot of really short (1 or 2 line) messages. I get some messages like that, but most of the ones I save wind up being 1 or 2K anyway, so I find that I save only about 15% in disk space by storing a large number of mail messages in one file rather than separately. Users who have 1000 messages that look like: From: joe@foo.bar.com Date: Tue, 10 Mar 92 00:00:00 GMT Subject: Lunch Let's have lunch tomorrow. -joe tend to save up to 30% by storing in one big file. Of course, they'd save a lot more by throwing away the mail headers, in this case. I dislike MH in general for several reasons I won't get into now, but I have to admit that the one-message-per-file philosophy has several advantages: 1. Easier manipulation of individual messages. 2. Much lower probability of catastophic mail loss in event of system crash or program bug which occurs while reading mail. Usually only one message at most will be lost, rather than entire folder. (This point can't be brushed aside easily; the manipulation of files larger than a megabyte triggers obscure bugs in several programs and filesystems.) 3. Programs reading such files can be written to use less memory than programs which have to hold entire folder in memory at once. This means reduced swapping and less likelihood of running out of memory on low-end or heavily-loaded systems. I would like to see VM be able to manipulate messages in some one-message-per-file format, especially if it is optional; i.e., I still would like to be able to read UNIX mbox-style mail folders, easily burst the messages into individual files, and get them back into one file, preferably all within VM. -Calvin -- Calvin Clark <ckclark@mit.edu>
Technical Details
Message-ID:
<CKCLARK.92Mar10202803@deathtongue.mit.edu>Path:
arc1-ca.pugleaf.net!sparky!uunet!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!usenet.coe.montana.edu!mintaka.lcs.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!ckclarkReferences:
<9203092235.AA04257@churchy.gnu.ai.mit.edu> <9203092237.AA28624@wendy-fate.UU.NET> <9203092247.AA07651@thalidomide.lucid> <9203092251.AA29857@wendy-fate.UU.NET> <9203092301.AA07757@thalidomide.lucid> <9203092306.AA01860@wendy-fate.UU.NET> <920309233Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: f6829,cfa18af4514eb905 X-Google-Attributes: gidf6829,public X-Google-ArrivalTime: 1992-03-10 17:52:41 PST Path: arc1-ca.pugleaf.net!sparky!uunet!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!usenet.coe.montana.edu!mintaka.lcs.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!ckclark From: ckclark@athena.mit.edu (Calvin Clark) Newsgroups: gnu.emacs.vm.bug Subject: Re: Babyl Message-ID: <CKCLARK.92Mar10202803@deathtongue.mit.edu> Date: 11 Mar 92 01:28:14 GMT References: <9203092235.AA04257@churchy.gnu.ai.mit.edu> <9203092237.AA28624@wendy-fate.UU.NET> <9203092247.AA07651@thalidomide.lucid> <9203092251.AA29857@wendy-fate.UU.NET> <9203092301.AA07757@thalidomide.lucid> <9203092306.AA01860@wendy-fate.UU.NET> <920309233 Sender: news@athena.mit.edu (News system) Reply-To: ckclark@mit.edu Organization: Massachusetts Institute of Technology Lines: 53 In-Reply-To: djm@eng.umd.edu's message of Tue, 10 Mar 92 00:47:08 GMT Nntp-Posting-Host: deathtongue.mit.edu