Comment 7 for bug 388269

The simple 1-revision search method does show more transfer for bzr, though;
23.269 Transferred: 2550kB (110.9kB/s r:1244kB w:1306kB)
15.020 Transferred: 3801kB (258.3kB/s r:1917kB w:1884kB)

That is the amount transferred before actual data streaming. Notice that it is quite a bit faster, but this is over my loopback.

The same requests to lp:bzr gives:
47.896 Transferred: 2545kB (54.5kB/s r:1242kB w:1303kB)
50.192 Transferred: 3796kB (78.0kB/s r:1915kB w:1881kB)

Note than now the extra data transfer actually makes them equally fast. Though I think we can do something smarter than just a single skipping revision. And certainly, *some* sort of limiter to avoid N^2 makes a lot of sense.