Ubuntu Domain Server

Derek Broughton derek at pointerstop.ca
Wed Oct 28 13:29:42 UTC 2009


Christopher Chan wrote:

> Derek Broughton wrote:
>>
>> All the RFCs are defined as finite-state engines.  There really is NO
>> reason that a tool capable of making all the correct configurations need
>> to be
>> "predefined" and "fixed".  It's 30 years since I did FSEs in university,
>> but I'm pretty sure we learned that they could _all_ be automated, even
>> then.
>>
>>   
> 
> Oh feel free to code the thing then. Just don't ask mom and pop whether
> they want their user account database in ldap or mysql or in passwd and
> shared via NIS+.

My whole point has been that it could be done, while you've been saying it 
couldn't.  Having apparently accepted that was wrong you raise spurious 
issues about implementation.  What does where they want their user accounts 
have to do with anything?  Pick a reasonable default and use it.  Ask them 
if they already have a user source, and use that.  If mom & pop are setting 
up an initial system, they'll happily use whatever you give them.

>> My recollection is that the "disk images" came after the initial
>> proposal,
>>   
> 
> Maybe you need to reread the first post then.

Not really.  So I missed it on first read, but as I said...
> 
>> but even so: "yeah".  What makes a _second_ disk image any more
>> significant
>> than the first?  If the first is correct, then the second, with specific
>> mods to make it reflect a unique machine, is not that difficult.
>>   
> Are still talking about mom and pop here? I imagined that they would get
> computers that come with UDS preinstalled? They are supposed to know
> what mods to make? Do I hear experienced professional required?

No, they aren't.  I just don't see the difficulty in having an automated 
system make the deltas.  You provide identical disk images of the fixed 
data, and each system gets its own image of the unique data.  Simple rocket 
science.
-- 
derek





More information about the Ubuntu-devel-discuss mailing list