This is about the function of difficulty in collaborative communities.
Existing tools were intimidating and built for the computer literate - new users exposed to raw data models and complex traditions were often lost at the start.
What we built is arguably friendlier and simpler, despite some quirks and bugs at the outset.
What we encountered in the process of 'launching iD' - making it the default editor for when people click 'edit' on the website - was interesting, because much of the feedback was not technical, but reflected a certain fear of open doors.
As others posted, I think the trash can icon is far too prominent: three clicks from the home page and an new contributor's first action with OSM could be to accidentally delete something! *
As long as the delete function is that prominent it is not an editor for newcomers. Please, do make it easier to use existing objects and do not lead the user to delete and then create a new object. *
A little while later, I received a note from Steven Walling that pointed out the similarity between our project and VisualEditor, a project by the WikiMedia Foundation to simplify editing Wikipedia.
Real-life feedback on the VisualEditor project.
And, within a few minutes, a now-deleted tweet from Gregory Kohs linked up his article in examiner.com about how the editor is potentially damaging the community's resource, because it makes it too easy for new users to make changes.
It's very difficult to build a safe editor for OpenStreetMap, just the same as it is to build a safe editor for Wikipedia. OSM may be 'just roads' and Wikipedia is just text.
But in the struggle to reflect the world's idiosyncracies, the data models of collaborative projects become twisted, fragile, and subtle in ways that work counter to the idea of collaboration.
Without eliminating these fault lines or having systems around them to prevent collapse, the community becomes paranoid about the damage that some unindoctrinated newcomer might bring on their first day of playing.
In OpenStreetMap, the best example is the idea of a 'relation'. While what you see on the map is made up of simple points and lines between them (nodes and ways in the lingo), relations form an invisible third layer - connecting boundaries, bus routes, multipolygons, turn restrictions, themselves, recursively, and much more.
Relations are non-geometric, non-visual entities with implied meaning that can be deleted carelessly by users who don't know their existence in the first place, and to try to explain them to newcomers is, let's say, tough.
The axis of Wikipedia is wiki markup, a non-standardized, expansive set of commands that connect pages and declare styles.
Markup languages are, as a species, divided into dialects - HTML into XHTML and HTML5, Markdown quickly splintering into extended and GitHub-flavored versions. The difficulty of representing information means that standards leak at the edges into trades of compatibility for power.
This happened with Wikitext - a standardized version called Creole failed to catch on to Wikipedia itself. And Wikipedia's version has expanded into relation-like complexity - interdependent templates form much of the structure of the encyclopedia, and those 'most active editors' actually spend much of their time refining style and dealing with 'special syntax', as discovered in Aaron Swartz's incredible Who Writes Wikipedia?.
The implication is that, as open data communities grow, the 'product' becomes larger and more complex - Wikipedia covers most topics you can think of and OpenStreetMap has billions of objects and thousands of high-profile users.
And the fragility of the project and the openness of the project mix: OSM has almost no permissions model, while Wikipedia has a few levels - nothing in comparison to the finely-tuned hierarchies of commercial sites, like Foursquare's Superuser statuses.
On the other side:
"You can edit 1 mile around roads you've driven". This is Waze's map editing interface.
"Your Edit Was Approved". This is Google Map Maker, with an unimportant change.
These 'access controls' look corporate and un-free from the other side, but in a way they are an explicit way to make sure that users have only as much power as they have knowledge and experience. It's an explicit implementation of the implicit partitioning that Wikipedia & OSM have with technical learning curves.
The annoying entry restrictions to open data projects and communities in general are not just technical, but community constructions. Changing them changes the community in a way that affects culture and content on every level.
For existing projects, this is going to be a struggle - the diversity of opinion, education, and intention that already causes mile-long flamewars is just a fraction of a true worldwide sample.
And going forward, it's important to consider how every technical choice will have a cultural effect, whether good or bad. Just like Linus chooses C to keep people out, what can we choose to bring people in?