The quantum internet holds promise for performing quantum communication, such as quantum teleportation and quantum key distribution, freely between any parties all over the globe. For such a quantum internet protocol, a general fundamental upper bound on the performance has been derived [K. Azuma, A. Mizutani, and H.-K. Lo, arXiv:1601.02933]. Here we consider its converse problem. In particular, we present a protocol constructible from any given quantum network, which is based on running quantum repeater schemes in parallel over the network. The performance of this protocol and the upper bound restrict the quantum capacity and the private capacity over the network from both sides. The optimality of the protocol is related to fundamental problems such as additivity questions for quantum channels and questions on the existence of a gap between quantum and private capacities.PACS numbers: 03.67. Hk, 03.67.Dd, 03.65.Ud, In the Internet, if a client communicates with a far distant client, the data travel across multiple networks. At present, the nodes and the communication channels in the networks are composed of physical devices governed by the laws of classical information theory, and the data flow obeys the celebrated max-flow min-cut theorem in graph theory. However, in the future, such classical nodes and channels should be replaced with quantum ones, whose network follows the rules of quantum information theory, rather than classical one. This network, called quantum internet, could accomplish tasks that are intractable in the realm of classical information processing, and it serves opportunities and challenges across a range of intellectual and technical frontiers, including quantum communication, computation, metrology, and simulation [1]. So far, the main interest in the quantum internet has been its realization [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. But, it must be one of the most fundamental trials from a theoretical perspective to grasp the full potential of the quantum internet. Along this line, recently, a general fundamental upper bound on the performance was derived [17] for its use for supplying two clients with entanglement or a secret key. Interestingly, this upper bound is estimable and applied to any private-key or entanglement distillation scheme that works over any network topology composed of arbitrary quantum channels by using arbitrary local operations and unlimited classical communication (LOCC). With this, for the case of linear lossy optical channel networks, it has been shown [17] that existing intercity quantum key distribution (QKD) protocols [18][19][20] and quantum repeater schemes [7,8,12,14,15] have no scaling gap with the fundamental upper bound. Moreover, in the case of a multipath network composed of a wide range of stretchable quantum channels (including lossy optical channels), it has been proven [21] to be optimal to choose a single path between two clients for running quantum repeater scheme, in order to minimize the number of times paths between them are used...