http://planet.i2p?i2paddresshelper=W0omtMApOdlTkgJoct08QnzDkeb~xc60df5Cx3k3qqOFmyEkRB65NdpSdvldI5STmXmXlZe68f8Hu1dfBMmKvfpBFMpP-mRVYuHbEa4N1dRkp3AOswZFO51lSC~MVPMZKGHAv0cfv5WFaJ9nfuaR7iYYWJF11xotoPTMbJdB-I3XPC8rxHI-73ATlOD4cU9oZ6BYcBE0dun9zw4IHBcRpnFRx~s-TkevNfPnBqujB-NNgA8a5s~QUj0h3FwwUWsNB8f1Uwy~DwoQh7pMp6dsZ3umEyBn2HaXZasSmbzB30qkQRs9vrAtIFdIzCQ~8gI2cHdHEPgExUQX~xlM8QZEgokoexilIPG53Co1kKARN0uCSxu8uCcP0FriGejPp3vdWdopxFLi2Oph4SMwKgD~QSLb54u-ZXlvBXLrKyr1xZC~KkDv45D4z8WN4kYjSpglcQGTEURWnYEDjVmS7oIrAiI7D-60cJ0qXfV2i7AjxmWaws4a41vV9hWJuRAzhZtYBQAEAAcAAA== rel=
Possible causes: - bugs - misreading of the Westwood+ papers - poor adaptation of the algorithm to our implementation which counts packets not bytes - bad ack strategy - poor choice of algorithm in our relatively high-latency, high-loss environment - bad test harnesses, giving us bad results and steering us in the wrong direction - poor interaction with vastly different implementations at the other end - tunnel breakage, with a possible fix documented in new proposal 171...