Path: utzoo!utgpu!cs.utexas.edu!wuarchive!zaphod.mps.ohio-state.edu!sdd.hp.com!ucsd!ucbvax!HANKEL.RUTGERS.EDU!bbs
From: bbs@HANKEL.RUTGERS.EDU (Trashy)
Newsgroups: alt.sources
Subject: Re: ELVIS WARNING - LOST CLUSTERS ON PC's
Message-ID: <9009032154.AA03937@hankel.rutgers.edu>
Date: 3 Sep 90 21:54:37 GMT
References: <1790@tuvie>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Society For Anachronistic Poetry, Central New Jersey Chapter
Lines: 32

In article <1790@tuvie>
	hp@vmars.tuwien.ac.at (Peter Holzer) writes:
]strobl@gmdzi.UUCP (Wolfgang Strobl) writes:
]>A really perfect argument. MSDOS filesystem semantics are different from
]>Unix filesystem semantics. This shows how poor DOS is. Oh well...
]
]If the filesystem semantics of MSDOS mean that you can remove (or rename)
]an open file and thereby corrupt your filesystem, that DOES show how poor
]DOS is, IMHO.
]
]If DOS is not able to defer the removal of an open file until the file is
]closed, it should not remove the file and return an error. Lost clusters 
]are not dangerous, but having your disk slowly filled with them is at least
]inconvenient.

This brings up a general point about the design of system services and
library software:  make them _robust_!  If you are writing a subroutine
for some particular application, you can afford to leave it with some
holes that you know the rest of the program is going to work around,
but as soon as your routine is meant for general application, then any
hole represents the potential for a plethora of bugged programs.

Why leave that potential in there?

MSDOS blows it big with the `rename' service call.  If _I_ had written
that I'd be embarrassed.



-- 
Barry Schwartz            bbs@hankel.rutgers.edu
mnetor!dciem!nrcaer!alzabo!trashman@uunet.uu.net