Elliot Murphy posted a comment on LWN clarifying his message (comment here: http://lwn.net/Articles/356947/ attached to parent http://lwn.net/Articles/356911/). The most relevant part is this: "It seems I was overenthusiastic when sending that email to the couchdb dev list, and left out an important word: 'can' be replicated :)"
In other words, this is an optional service that you can choose to use -- no default install is going to send your info to Canonical.
Id stick my neck out to say that something like CouchDB will become the de-facto standard for almost all data. Data that isn't {securely} on the web is in effect data that doesn't exist, while more data than not is naturally in a graph or tree-like form.
Mix in offline mode, JavaScript on client and server and you have a really nice development stack to make apps with.
I can imagine a generation of teenagers developing web apps [they'll think there is no other kind of app] using map / reduce idioms to get to their data, SQL a thing of the past.
Myself I'm also into document DBs and graph DBs, a lot. But for the kind of rapid prototyping that got most of us going as teenagers in the first place, I'd argue table based data with the simplest incarnations of INSERT UPDATE DELETE SELECT WHERE haven't been beaten yet by those no-SQL DBs.
This is an attractive feature of CouchDB that I'm considering for a web application too.
I guess a link works well enough.
