Subject: Re: inn and mmap()ing the active file
To: None <greywolf@tomcat.VAS.viewlogic.com, toor@dyson.iquest.net>
From: Captech) <greywolf@tomcat.VAS.viewlogic.com (James Graham>
List: current-users
Date: 08/11/1995 10:13:02
#: From: "John S. Dyson" <toor@dyson.iquest.net>
#: Subject: Re: inn and mmap()ing the active file
#:
#: > Indeed.  Anyone who is using the mmap() stuff without using msync() is
#: > a) asking for trouble and b) doesn't know how to use mmap() properly and
#: 
#: Only on early *BSDs, on FreeBSD that is the reason for the ports tree.
#: Either when the original author does not directly support FreeBSD or when
#: FreeBSD does something wrong in some release or other...
#: 
#: > should therefore RTFM before playing with his new toy.
#: > 
#: > Who *is* the author of INN anyway?
#: > 
#: Actually it is a buggy system that needs the msync.  This is the reason
#: for the FreeBSD merged VM/Buffer Cache monster bug-fix.  Yucky SVR4 even does it
#: correctly!!!  (I used to maintain SVR4 for a major computer vendor and
#: it is amazing that it does work!!!)
#: 

Maybe we're talking about different things here, then.  SunOS must also
be buggy by your definition, because it states that neither close()
nor fsync() will have an effect on an mmap()ed file -- one must use
msync().

#: John
#: dyson@freebsd.org
#: 
#: