[AFS3-std] AFS Standardisation

[Jeffrey Hutzelman]

--On Wednesday, August 20, 2008 12:09:19 PM -0400 Jake Thebault-Spieker 
<summatusmentis@gmail.com> wrote:

>> The Registrars are specified in the proposal as:
>> "The registrars would be initially comprised of the current registrar,
>> plus a representative from each of the currently available AFS
>> implementations (IBM, OpenAFS, kAFS and Arla)"
>>
>
> Actually, valid point raised above. Suppose there aren't representatives
> available/willing/able to represent one or more of the current AFS
> implementations?

That's not actually a problem.  The registrar's job is to maintain the 
lists of what values are assigned for what purpose, to insure that the 
necessary information is provided when a value is assigned, and to serve as 
the synchronization point that is needed to avoid having the same value 
assigned for more than one purpose.  The role specifically does not involve 
passing judgement on whether the intended purpose of a value is a good 
idea; that is left up to the standardization group to determine, either by 
endorsing a standard for which a value is to be assigned or by having 
created a registry in which values may be assigned without a standard.

The registrar load is actually light enough that one person can easily 
handle it.  The reason for having multiple registrars is for redundancy to 
protect against one person becoming unavailable or unwilling to serve or 
too lame to respond in a reasonable time.  Once the group is set up, it is 
expected to be self-maintaining, selecting new people as needed.  The 
initial bootstrapping could just as easily be done by making the initial 
group be composed of the current registrar (i.e. me) and 2-4 people of his 
choosing.  Simon can correct me if I'm wrong, but I suspect the reason for 
starting with representatives of the existing implementations was related 
to the role of the Registrar in the initial election.

-- Jeff