Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!mimsy!chris
From: chris@mimsy.UUCP (Chris Torek)
Newsgroups: comp.unix.wizards
Subject: Re: symbolic links poster child
Message-ID: <7326@mimsy.UUCP>
Date: Sat, 4-Jul-87 03:17:03 EDT
Article-I.D.: mimsy.7326
Posted: Sat Jul  4 03:17:03 1987
Date-Received: Sat, 4-Jul-87 17:28:34 EDT
References: <7955@brl-adm.ARPA> <2249@bunker.UUCP> <244@nuchat.UUCP>
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 25

>In article <2249@bunker.UUCP> zink@bunker.UUCP (David Zink) writes:
>>I guess that my major complaint is that ELOOP purports to mean LOOP but
>>really means EIGHT.

(Would you prefer ETWELVE? :-) )

In article <244@nuchat.UUCP> steve@nuchat.UUCP (Steve Nuchia) writes:
>For reference, if anyone is in a position to fix this, there is a very
>good loop detection algorithm that takes 1.5 X the time and 2 X the space
>of the basic linear search it protects. ...

Of course, the present `loop detectection' takes 1 X the time and
1 (no multiplier) space unit.  Since namei accounts for most of
the time spent in the kernel, slowing it even a little bit may be
too much.

Moreover, eight symlinks is really not all that small a number.
We had five levels of symlinks in one translation on a Sun, and it
was noticeably slower than other path name translations.  If you
have enough levels of symlinks that you run into ELOOP errors, you
probably have too complex a nest of symlinks.  (That said, I do
think eight is too small; but I also think namei is too slow.)
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690)
Domain:	chris@mimsy.umd.edu	Path:	seismo!mimsy!chris