Subject: Re: netatalk in production environments....(RFC)
From: Sak Wathanasin (sw@nan.co.uk)
Date: Thu Jun 01 2000 - 07:01:51 EDT
In reply to Ron Chmara's message of the 01/06/2000 at 01:27 -0700,
>More to the point, a pre-press house I used to work in once blew $500 of
>bad film, matchprint, *and* missed a press date, because an alias was
>pointing to a folder that supposedly held the right data. It did not,
>it pointed to an *older* version. This almost cost the company a $120K
>client.... hence, their standing rule to _only_ access data by a direct
>path, because if the user _had_ gone through the tree, they would have
>seen the other folder, titled "•••NEW VERSION HERE!!!!!!".
It seems to me that it wasn't the alias that was the problem, it was
caused by whoever created the new version. The thing to do there,
IMHO, was to dup the original folder, rename it to "copy of the
mm/dd/yyyy" and put the up-to-date contents where all the aliases
were pointing. Same as you would do on a Unix system with sym-links.
Rather than ban aliases outright (is making users drill down through
a dozen or so links that less error-prone?), you could use a chained
alias. Instead of creating aliases to the final target directly,
point your aliases (on the users' disks) at an intermediate alias
called something like "current version" on the server. Then when the
target moves, you only have to change 1 alias. Again just like you'd
use sym-links on Unix.
It isn't just the "dumb" users that are the problem; they are under
immense pressure to get stuff out of the door, and they WILL take any
shortcuts available to them; it's up to us to make sure that the
shortcuts work properly.
-- Sak Wathanasin Network Analysis Limited 178 Wainbody Ave South, Coventry CV3 6BX, UKInternet: sw@nan.co.uk Phone: (+44) 24 76 41 99 96 Mobile: (+44) 79 70 75 19 12 Fax: (+44) 24 76 69 06 90
This archive was generated by hypermail 2b28 : Wed Jan 17 2001 - 14:30:52 EST