Colocation between observing stations of the various different space geodetic techniques is critically important. It is the usual and most straightforward means of unifying the separate, independent realizations of terrestrial reference frames to establish the ITRF. Colocation also makes it possible to perform direct comparisons of geodetic results to assess technique-dependent systematic errors. Unfortunately, the overall accuracy of the current ITRF appears to be about the same level as the one-day precision being achieved with the best observing techniques and data sets. This means that little or no accuracy improvement is being gained from long-term integration for such quantities as station coordinates. Most likely, the benefits of integration are limited by sparseness of colocation sites, by inadequate local tie information, and by neglected systematic errors. Improvements in all these respects are needed.