Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!crdgw1!uakari.primate.wisc.edu!sdd.hp.com!news.cs.indiana.edu!att!ucbvax!FTP.COM!backman
From: backman@FTP.COM (Larry Backman)
Newsgroups: comp.protocols.tcp-ip.ibmpc
Subject: Re: Wildcards turn to ALL-CAPS file names on IFS
Message-ID: <9105201409.AA12922@ftp.com>
Date: 20 May 91 14:09:28 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 50


 >> 
 >> 
 >> I'm using FTP's PCTCP for OS/2 1.1 with NFS mounted disks from
 >> a Unix box.  The FTP documentation indicates that when using
 >> wild cards to specify files names (eg. copy *.*), the resulting
 >> file names returned from the operating system (presumably by
 >> DosFindFirst/DosFindNext) are always uppercased.  This is a
 >> real pain on the Unix side where lower case is the rule.
 >> 
Yes; this is a pain.  It leaves a mess behind & there is no way around
it.  I've considered adding a 
	case= {LC | UC}
switch to idrive.ini to make life easier from the UNIX side.  This of
course is a performance hit however, as each time a file is opened its
has to be case checked (what if you really do have a file.TXT?) & case
swapped as needed.

 >> Has Microsoft or IBM considered the effect of case on
 >> installable file systems?  This should affect HPFS as well,
 >> although not quite so much since HPFS is case sensitive only
 >> when creating a file name, not when reading or matching one.
 >>  
They have, I had many long arguments with them last year, hopefully
some of it sunk in.  Unfortunenately; it seems they never considered
Unix & NFS in their plans, and made the grand assumption that they controlled
the namespace of any file systems.

Just for your information, the case issue is a *real* performance drag.
Whenever a filename lookup fails; NFS has to go back & read the entire
directory & serach for a case-insensitiove match.  This adds many extra
packets to your normal transactions.  for instance: if you have a file:

	file.TXT

& you call it up in your editor:

	e file.txt

the initial lookup of file.txt fails; NFS goes back & reads the entire 
directory ( at least 2-4 extra packets!) finds file.TXT and loads it.

The worst part of it is that the semi-case sensitivity wreaks havoc with
any attempt to maintain a FileName/Handle/Attribute cache.  I've got
tons of space in OS/2 to maintain such a cache; but half the time I
can't trust it due to case {in}sensitivity issues.



						Larry