home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!dtix!darwin.sura.net!uvaarpa!concert!duke!trt
- From: trt@duke.cs.duke.edu (Tom Truscott)
- Newsgroups: comp.unix.aix
- Subject: Re: documentation on getpriv/setpriv?
- Message-ID: <713665972@romeo.cs.duke.edu>
- Date: 13 Aug 92 00:32:53 GMT
- References: <1992Aug5.195224.27457@awdprime.austin.ibm.com> <LWVANELS.92Aug8190323@fionavar.mit.edu> <1992Aug9.190521.21579@awdprime.austin.ibm.com>
- Organization: IBM RTP
- Lines: 15
-
- >It is quite easy to write
- >code which uses the access() and other functions.
-
- But access() cannot be used to provide security, because of
- race problems for example.
- Several CERT advisories have been due to use of this call.
- Indeed there are virtually no reasonable correct uses of this call.
- I have a thorough discussion of this, send mail if you want a copy.
-
- The man page for access() should discourage its use.
- There are other calls and programming practices that should be avoided
- as well. A "SPECIAL CONSIDERATIONS" section of the man page
- could do this for access(), gets(), O_NSHARE and O_RSHARE, and so on
- (setpriv/getpriv?). It is important to be able to take this first step,
- because without it no "feature" can ever be eliminated.
-