Love it. FWIW, here are some thoughts from Build CouchDB feedback. On Fri, Nov 4, 2011 at 5:26 AM, Jan Lehnardt <j...@apache.org> wrote: > Randall, Till, > > I had hoped I addressed this in my original mail, but let me try again :) > > I'm in favour of all the things your are saying, but I'm trying to address > this scenario: > > $ wget $url > $ tar xzf $archive > $ ./$dir/couchdb > Time to relax.
Basically, I suggest clearly identifying every non-goal early. It will help to assess exactly what people want. I am not sure ./bin/couchdb is what people want. I've seen disappointment from build-couchdb users. The world has changed since Noah ported CouchDB to autotools. System administrators are an endangered species. Time was, once you got `make install` working, it was downhill from there. Now, people don't find that as relaxing. * If there is a system couch, you get a long scary message with 'eaddrinuse' buried somewhere in it * When you reboot or computer crashes (possibly 6 months later), Couch isn't running anymore * Log rotation doesn't work (eventually breaking the /_log query in version =< 1.1.1) * It's unclear how a backup/restore would work. * It's unclear how an upgrade would work. * It's only "native" in the strictest sense, you can't manage it with the OS (i.e. it's not an .app bundle, .msi installer, .pkg, .rpm etc.) The first email I got about Build CouchDB, somebody said thanks and that it helped. Not one hour later, two emails complained that there is no Ubuntu package. It was really discouraging. But I'd done poor expectation management. So, the above may be excessive for now, but at least clearly describe goals and non-goals. -- Iris Couch
