[OSM-dev] The future of sysadmin

[Lars Aronsson]

Steve Coast wrote:

> I'm still accused of centralising control and being  
> evil (I'm looking at you Lars).

I like it when you write long messages, Steve, but it happens far 
too seldom.  You should make it a habit.  All too often I only got 
a one-liner from you, and your "have fun" .signature which I 
always found a bit arrogant.  Not only did that add to whatever 
frustration I was carrying, it also kept me from asking again.  
We need to communicate more, not less.  My whining would be 
totally without merit if I didn't contribute to the project from 
time to time.

I can't travel to London to visit the servers, so I hope you find 
someone who can.  In my experience, responsibilities can be 
distributed.  There were many cases where a simple "ssh" and a 
database restart could have helped OSM, but nobody was trusted to 
do this.  Many could help with that (using setuid or sudo), 
without travelling to London and without "getting root".

For some time this spring, I did put great hopes in the tiles at home 
project, but these have faded.  I'm now more inclined to 
Frederik's ideas of live feeds.  That change could transform the 
project towards a more distributed architecture, very much like 
the introduction of the planet.osm dumps did.  In any such 
solution we must make sure that the maximum amount of work can be 
distributed, and the remaining central component is as small as 
possible.  This is where I think t at h has failed: Too much work 
(index updates) is still centralized, and there is no obvious way 
to move that work from central server to distributed clients.

> How have other projects managed it?

This computer room, with a mix of Solaris, BSD, Linux and more,
http://www.lysator.liu.se/media/foo_large.jpg
http://images.google.com/images?q=foo-hallen

belongs to the students' computer club http://www.lysator.liu.se/

and is operated by the club's "root team", mostly undergrads, lead 
by a single person titled "super root", all volunteers.  I have 
just a small server in there (runeberg.org) where we are only 
three people with root access. I have a key to the computer room, 
but am not a member of the club's root team.  The club was founded 
in 1973, 
http://www.lysator.liu.se/history/images/D21-invigning-konsoll.gif

and has had Unix servers on the Net since 1989.  This computer 
room has been used since 1996.


-- 
  Lars Aronsson (lars at aronsson.se)
  Aronsson Datateknik - http://aronsson.se