[OpenAFS-devel] Re: setpag switch for afslog?

Russ Allbery rra@stanford.edu
Mon, 23 Feb 2004 21:16:08 -0800


Jeffrey Hutzelman <jhutz@cmu.edu> writes:

> That "feature" is a horrible kludge.  It severely breaks a fundamental
> abstraction (inheritance is one-way), and like Love, I'm not totally
> convinced it's secure.  It should never have been there in the first
> place. Speaking as an individual, I strongly encourage folks not to use
> this functionality or write code that depends on it.

> That said, a lot of people are depending on this, so I expect it will be
> around for a while.

I was, amusingly, completely unaware of its existence until about a year
ago and have never used it, despite having used and then maintained AFS
for more than ten years now, so I can attest that it's definitely not
necessary.

One develops somewhat more elaborate ways of invoking things inside pagsh
and then acquiring tokens and then scripts them, and then things are
basically as convenient.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>