Path: utzoo!telecom-request Date: Thu, 2 May 91 07:53:40 PDT From: "Louis J. Judice 02-May-1991 1033" Newsgroups: comp.dcom.telecom Subject: Re: Prodigy or Fraudigy Message-ID: Organization: TELECOM Digest Sender: Telecom@eecs.nwu.edu Approved: Telecom@eecs.nwu.edu X-Submissions-To: telecom@eecs.nwu.edu X-Administrivia-To: telecom-request@eecs.nwu.edu X-Telecom-Digest: Volume 11, Issue 329, Message 2 of 9 Lines: 26 > I know its off the topic, but ... if you are on a multi-user system > and this technique works for you ... switch. That is terrible > security and the vendor deserves not to be in business (don't name > names, I know several which work this way). Since most of our > multi-user readers are on UNIX, this trick will not work on UNIX > systems. > Two reasons: First, UNIX does not allocate the intervening > space in the file. It just allocates the blocks you write to. The OS > returns 0's for all other blocks read that are not yet allocated. > Second, UNIX does not write partial sectors, nor depend on the > contents of the file to mark end of file. Not to stray even further off the topic, but I would caution readers that in a production environment, this "feature" would probably end up requiring a lot more I/O, especially on a fragmented disk. Your best bet is to use file highwater marking techniques on timesharing disks and permit preallocated files on disks holding large production databases. My guess is that most big UNIX database systems work the latter way, using raw I/O. And of course other operating systems, like VMS, also let you "go both ways." :) ljj