[325] in Info-AFS_Redistribution
Re: the dangers of not using unlog
daemon@ATHENA.MIT.EDU (Bradley White)
Fri Sep 20 16:24:26 1991
To: info-afs@transarc.com
In-Reply-To: Bob Andrews' message of Fri, 20 Sep 91 13:17:58 -0400
Date: Fri, 20 Sep 91 15:14:12 -0400 (EDT)
From: Bradley White <bww+@cs.cmu.edu>
bob@ibmpa.awdpa.ibm.com writes:
> reuss@ni.umd.edu writes:
>>
>> This 10 second limitation obviously means you can only expect 6
>> authenticated connections to a machines per minutes, but in reality I found
>> it was even less than this. ...
>>
>> -Karl Reuss
>
> I think ten seconds is a little excessive. One second might be better. ...
>
> Bob
Any non-conditional delay is excessive, particularly given that the
current scheme does not even *guarantee* that a PAG will never be
reused (no matter how many maximum-uptime statistics you can quote).
A simple optimistic method which keeps track of the range of active
PAGs would mean that typically there would be no delay. And even in
the unlikely event of a possible reuse collision, checking the PAGs
actually in use would undoubtedly be both faster and correct.
Bradley
