Path: utzoo!attcan!uunet!cs.utexas.edu!execu!sequoia!cb
From: cb@sequoia.execu.com (Christopher D. Brown)
Newsgroups: comp.os.msdos.programmer
Subject: Re: Implementation of File Locking
Message-ID: <22586@sequoia.execu.com>
Date: 6 Aug 90 13:13:52 GMT
References: <1990Aug3.182050.8793@ns.network.com>
Organization: Execucom Systems Corp.
Lines: 28

In article <1990Aug3.182050.8793@ns.network.com> dmb@ns.network.com (Duane M. Butler) writes:
[preamble deleted]
>One idea I'm thinking about is to create a temporary lock file in the directory
>where the files are.  Each time my program accesses the files in the directory,
>it first checks for the existence of this lock file.  If present, it will try
>again in 10 seconds or so, and give up after 3 trys.  When the program attempts
>to modify the control files, an attempt is made to create a lock file.  What I
>don't know is if the create file operation is atomic.  In other words, is it
>possible that two instances of the program could believe they created the lock
>file at the same time?
>
>If anyone has solved this problem I would appreciate hearing from you.

PVCS, built by Polytron now from Sage, uses directory creation rather
than file creation to implement semiphors.  I do not know if they had
a reason or viewed the choice as arbitary.

BTW.  As much as I feel that available standards (e.g. MS-DOS locking
via SHARE) ought to be used, in this case your choice is the only
one that yields reliable coverage of a wide variety of environments.

cb
-- 
Christopher D. Brown

Digital: {uunet|texbell|cs.utexas.edu}!execu!cb
Analog: (512) 327-7070
Physical: Execucom, 108 Wild Basin Road, Two Wild Basin, Austin, TX 78764