[OSM-talk] Big sponsors

Frederik Ramm frederik at remote.org
Fri Jun 18 08:19:35 BST 2010


Hi,

SteveC wrote:
> Really it's the API we're talking about. Tiles are just a CDN
> problem.

In that case:

1. Actually allowing many people a day to make edits is not hard.

Your example was a site visited by 300 million people a day. If 1% of 
these go on to OSM and another 10% of those edit something, that's 300k 
people editing something. Not being power users, they'd just make a few 
edits, let's say 10 each. All these numbers are on the high side. That's 
3 million edits a day. We had 1.5 million yesterday - so it's not really 
an order of magnitude we're talking about, it's just Grant plugging in a 
few SSDs and then, bring them on.

2. The _technical_ problem are the associated _read_ requests.

That is, people downloading an area before editing; people loading the 
Map Features page on the wiki, that kind of thing. These can easily be 
offloaded by a CDN (in terms of wiki etc.) and read-only database 
mirrors. It is then a small step to allow editors to read from mirror, 
post to central API. It's a SMOP. And those CDN/mirrors can be operated 
by Google or Microsoft for all I care. It is also, just like the tiles 
situation, never an exclusive partnership; anyone else can run the same 
infrastructure if they feel the need.

Only thing I would be religious about is not letting those partners have 
our user information ("sign up here with your hotmail/gmail/whatever 
account to start editing OpenStreetMap") - that would be one step too 
far, the community is _our_ community and not theirs.

3. The _community_ problem is that not every feedback is worth having.

If someone can be bothered to edit the map through any of the available 
editors, and that editor is not completely broken, that's a win.

If someone cannot edit but is capable of basic thinking, reading, and 
writing skills, they can place a meaningful feedback on OpenStreetBugs 
or some other platform.

However there are many people who are unable or too lazy to deploy the 
aforementioned skills. A feedback like "Macy's missing on 3rd" is not 
feedback, it is rubbish. (What is Macy's? Where on 3rd is it? Is the 
marker even remotely in the right place? And is a plain and possibly 
special-interest POI really worth 15 minutes of mapper time to chase it 
up?) If linked from a popular site that attracts a broad section of the 
population, we will get lots of this kind of feedback. We don't 
currently have a 1st/2nd level support structure - our mappers do all 
the work. They cannot be burdened additionally with picking raisings out 
of rubbish.

So a feedback system for that kind of job would require (a) very good UI 
and docs to allow people to give non-rubbish feedback; (b) a way for 
voluntary 1st level supporters to process incoming feedback and sort it 
into action queues ("rubbish", "needs mapper to check/enter data", 
"needs someone to check facts locally", ...), and most of all (c) a few 
thousand voluntary 1st level support staff.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"




More information about the talk mailing list