Xref: utzoo alt.sys.sun:2268 comp.windows.x:30321 Path: utzoo!utgpu!cs.utexas.edu!sdd.hp.com!usc!apple!olivea!tymix!cirrusl!sunstorm!dhesi From: dhesi%cirrusl@oliveb.ATC.olivetti.com (Rahul Dhesi) Newsgroups: alt.sys.sun,comp.windows.x Subject: Re: /etc/protocols (was Re: Xkernel can't find NFS mounted directories) Message-ID: <2783@cirrusl.UUCP> Date: 7 Dec 90 04:43:30 GMT References: <1990Dec5.225824.10377@melba.bby.oz.au> <1990Dec6.052343.22445@Think.COM> Sender: news@cirrusl.UUCP Followup-To: alt.sys.sun Organization: Cirrus Logic Inc. Lines: 22 >>Now, why in EARTH put that in a file? IS sun afraid that the protocol >>numbers are about to change.....???????? The /etc/protocols file originated with BSD networking. At that time, I'm sure which protocols would exist in the future was very uncertain. (Some might claim that it's even more uncertain today.) According to Nemeth, Snyder, and Seebass (Unix System Administration Handbook, p 264), "if a user writes his own protocol using the raw socket interface provided for such purposes, it will have to be listed in the /etc/protocols file in order for inetd to manage daemons using it." So long as we allow protocols to be added to /etc/protocols, it makes sense to put all of them there. Perhaps the real question is, why sequential access of a text file rather than quick look-up of a direct access database. Perhaps the answer is that the file is small and read infrequently (no more than once by a process), and database style access is available, at least on some machines, through YP/NIS. -- Rahul Dhesi UUCP: oliveb!cirrusl!dhesi