There are two areas related to Internet measurement which are of particular interest to me and to my employers. As the use of communicating WWW caching servers becomes more widespread, the need increases for a reasonably intelligent means for a server to select the other cache or caches, if any, through which objects are to be retrieved. This requires the development of heuristics for predicting the performance of particular links or combinations of links, a problem very closely related to the measurement of congestion. Of more general interest to my employers is the development of means for the distribution of information about load and other conditions which affect network performance. It seems likely that progress in this second area will also be helpful in work on my own more specific interest, but it would also be capable of much wider application.