Currently, the code which handles support of the new depth stuffs (--depth, sparse directories, etc.) when talking to a pre-1.5 (non-depth-aware) server does so by simply throwing away information retrieved from the server that falls outside the scope of the requested depth. This is an accurate way to deal with the problem, but isn't optimal for RA layers which use multiple requests to get the job done. For example, ra_serf requests a skeletal tree delta report from the server, then subsequently (and somewhat simultaneously) fetches file contents and such as necessary. Today, ra_serf behaves exactly the same with the depth-filtering code in place as without. But ideally, it wouldn't even do those extra file content fetches for files that fall outside the depth scope -- those file contents are today immediately discarded anyway.