[OSM-dev] Alternative tile webserver needed?
frederik at remote.org
Mon Apr 30 00:36:19 BST 2007
> Yes, this is a stupid idea (no offense, the idea is stupid, not
> you) because it increases complexity enormously and makes the
> over-all system harder to understand and harder to maintain.
In our case, I really wouldn't mind if our servers were distributed and
our data split into regional parts. Our aim is to have free geodata for
the whole world; we don't need it on one server or even in one coherent
dataset. If I ever want to make a database request that returns data
from Finland and South Korea at the same time, I could live with this
request being more complex and maybe having to ask multiple servers.
Operating an OpenStreetMap database server needs, among other things,
manpower and money. Both may be easier to obtain locally, i.e. I could
imagine that it might be easier to persuade some entity in Japan to
commit servers and people who run them if they get the reward of being
able to say: "We run Japan's OpenStreetMap". (Of course they don't have
to call themselves OpenStreetMap either.)
I don't have any such reservations myself, but I believe in many parts
of the world, people would rather like to contribute to a "local"
project than to something run by imperialist Europeans (or even worse,
Americans), especially if it is to do with the map of the country they
> For a while,
> somebody sponsored a separate server in Korea for the Wikipedias
> in Japanese and Korean. But coordinating the sysadmin people in
> both places was too hard, so they moved everything back to one
> place in Florida (in September 2006).
I cannot speak for Wikipedia, but for us, I don't see why we would have
to coordinate sysadmin people at all. If the Japan OSM was completely
separate from the rest, using more or less the same software (but not
necessarily being "in sync", i.e. if we upgrade to API 0.8 and they're
still at 0.7, who cares - I know that you, Lars, are an exception but
99,9% of users will not feel the need to make edits on half the globe).
Of course sysadmins would likely want to share their knowledge - but
they don't have to.
I think the complexity - that leads you to call the idea "clueless" -
comes in if one aims at having the whole system behave as one,
distributing stuff automatically, making it accessible through one
central interface and so on. But we don't need that.
I am a technical person. I would prefer one big central database from
which mirrors get their updates, and if it works well, it can be located
in North Korea for all I care (would be a bit tough for Nick probably).
But I can see the point for regional servers - robustness, distributed
load, a better feeling of responsibility by the local admins, more
sponsorship options, plus the avoidance of the "imperialist" taste.
Your comparison with the canary bird doesn't work for me, because you
imply that the call for regional servers only comes up if something is
not working properly with the central ones; I can see some points that
can never be "fixed".
Frederik Ramm ## eMail frederik at remote.org ## N49°00.09' E008°23.33'
More information about the dev