Return-Path: wpp@marie.physik.tu-berlin.de Received: from raven.native-ed.bc.ca (raven.native-ed.bc.ca [134.87.106.1]) by nobozo.CS.Berkeley.EDU (8.6.4/8.6.3) with ESMTP id DAA08503 for ; Thu, 12 May 1994 03:23:13 -0700 Received: from mailgzrz.TU-Berlin.DE (mailgzrz.TU-Berlin.DE [130.149.4.10]) by raven.native-ed.bc.ca (8.6.4/8.6.4) with SMTP id CAA25098 for ; Thu, 12 May 1994 02:33:52 -0700 Received: from marie.physik.TU-Berlin.DE by mailgzrz.TU-Berlin.DE (5.65c/ZRZ-MX) for id AA21948; Thu, 12 May 1994 09:24:50 +0200 Received: by marie.physik.tu-berlin.de (5.0/SMI-SVR4) id AA01666; Thu, 12 May 1994 09:21:47 --100 From: wpp@marie.physik.tu-berlin.de (Kai Petzke) Message-Id: <9405120721.AA01666@marie.physik.tu-berlin.de> Subject: user authentification To: linux-postgres@native-ed.bc.ca Date: Wed, 11 May 1994 22:29:45 +0200 (MET DST) X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 1280 Hi, I am looking for a small project, which to start with hacking postgres. I do not want to do the big "C++"-ifying and "Web"bing thing, before they have released the final version. My suggestions are: - Modify the copy in/out routines to adapt them to a variety of input or output file formats. The current implementation takes about 800 lines in one source file: ~/src/backend/commands/copy.c - Add medium security authentification to postgres. Currently, you have no security (everybody can connect to port 4321, while a postmaster is running), or good security, when you link in Kerberos. Kerberos needs an independant ticket server, which should run on a physically safe computer, which has no other stuff running. Installing Kerberos requires you to change the login software. I want something in between, which provides both good safety and is easy to install. How about doing the same thing, that Oracle does: an extra login when connecting to the database? The problem: Packet Sniffer. While transferring the password, anybody can listen. So all data transferred during authentification should be encrypted. I came to mind with a strange scheme, how this could be done. I have written a post to sci.crypt about it. Kai