Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!seismo!epiwrl!epimass!jbuck
From: jbuck@epimass.UUCP
Newsgroups: comp.os.eunice
Subject: Re: vms format files and unix format files
Message-ID: <999@epimass.UUCP>
Date: Tue, 24-Mar-87 18:40:33 EST
Article-I.D.: epimass.999
Posted: Tue Mar 24 18:40:33 1987
Date-Received: Thu, 26-Mar-87 04:44:27 EST
References: <1250@uwmacc.UUCP> <17989@ucbvax.BERKELEY.EDU>
Reply-To: jbuck@epimass.UUCP (Joe Buck)
Organization: Entropic Processing, Inc., Cupertino, CA
Lines: 38

In article <17989@ucbvax.BERKELEY.EDU> leres@ucbarpa.Berkeley.EDU (Craig Leres) writes:

>There's a reason why exec() works the way it does under Eunice; it's
>faster for the 99% case. If the file is fixed length, exec() simply
>tries to execute it as a VMS image. If it's a variable length file, it
>reads a record to extract the name of the shell to use.

Ah, but it would add no overhead at all to see if it's a shell script
after trying to execute it as a VMS image fails.

If the Eunice developers were so concerned about speed they should
have made sure that scripts beginning with

#! /bin/csh -f

work correctly (the -f flag is ignored, meaning the user's .cshrc
file is always read).

>I disagree with the statement that random seeks only work with Unix
>files; although not efficient, random seeks work fine. 

You misquoted (or rather, misparaphrased) me.  I said you can't
UPDATE a "VMS format" file using random seeks (as in read-modify-write).
If you try, you'll get garbage.

>They are slow
>when dealing with a variable length file is because you must seek to
>the beginning of the file and read until you get to the right place.
>So make sure /etc/termcap is a fixed length file!

Doesn't help.  /etc/termcap is read from the beginning.



-- 
- Joe Buck 	{hplabs,ihnp4,sun,ames}!oliveb!epimass!jbuck
		seismo!epiwrl!epimass!jbuck  {pesnta,tymix,apple}!epimass!jbuck
  Entropic Processing, Inc., Cupertino, California