I really like this article. It's refreshing in that it doesn't yammer on about MOND Saving the Day™ while remaining healthily sceptical and informative.
Recalling one of the multi-week problems we were assigned during the General Relativity II, you can reasonably easily extend GR to derive a model of the universe without dark matter by 'simply' assuming that the flow of time measured by an observer has to be indistinguishable from a locally flat universe. When you check for (rotational) velocity dispersion function in such a model, it will be consistent with observations, provided you consider motion with respect to the CMB. However, as a consequence, it will turn out that the age of our universe is a couple of billion years older (16 or 17 IIRC, it depends on energy-mass density), which – to my knowledge – isn't easy to reconcile with other observations.
There's not much point to those ramblings, however, problems like those make me very happy I went for condensed matter instead of cosmology. :P