HN 提问：有没有人在没有 DHT 性能下降的情况下，解决了超过 1000 亿节点的 P2P 路由问题？
2 分•作者: ikhrabry•大约 13 小时前
我们两人（一位物理学和数学博士，一位哲学人类学家）多年来一直致力于研究一种数学上安全的秘密投票系统，适用于做出具有社会意义的决策。在设计过程中，我们意识到这样的系统必须是一个满足六个绝对要求的点对点网络。
我们未能找到任何一个现有的协议或学术论文能够同时满足其中两个以上的要求：
第一：匿名的用户身份验证，它保证用户在网络中表示的唯一性（防止一个人创建多个账户），同时不存储其个人数据，也不使用任何外部资源，例如证书颁发机构或可信第三方。
第二：提供此类匿名身份验证的网络必须能够覆盖地球上的所有居民。目前人口略高于80亿，但每年仍增长超过1亿。此外，已故者也不能被排除在系统之外，这意味着单调增长是无止境的。
第三：保证匿名的网络不能有任何特殊的节点——也就是说，它必须是严格的点对点网络，由用户设备组成的节点构成。然而，每个用户可能拥有不止一个设备，设备可能会发生故障、丢失、过时，甚至仅仅是过时。这立即改变了网络的规模——它必须处理数百亿个节点。
第四：有必要为每个用户提供从任何设备访问其账户的权限，而无需使用密码、令牌、助记词或任何类似的东西。
第五：关键要求是执行一个搜索，能够在1秒内完成用户识别（使其对用户不可察觉）。也就是说，任何设备必须在1秒内定位任何其他设备或数据，同时单个设备存储的路由信息量必须微乎其微，最好低于100 KB。
第六：为了保证不可靠的用户设备上的数据保存，首先，复制因子必须约为100倍甚至1000倍，其次，必须在没有中央仲裁者或可信节点的情况下，基于共识进行路由和数据的验证与恢复。同时，后台同步流量必须保持在每个设备每秒几百字节以下，以避免耗尽移动数据套餐。
我们的问题是：
社区是否知道任何现有的点对点协议、DHT或学术提案，能够尝试同时满足这些约束条件？我们研究了Kademlia、Chord、BATMAN、GNUnet、I2P等，它们似乎都在可扩展性或流量效率方面存在不足。是我们遗漏了什么根本性的东西，还是这确实是一个盲点？
（这个问题是基于我们已经开发了一个候选架构，在我们看来，它解决了所有六个问题。简要描述可以在我们的GitHub上找到 [https://github.com/ikhrabry-spec/A-P2P-Network-Architecture-for-100B-Nodes]。我们已经提交了专利申请，并获得了一项专利。然而，我们怀疑我们解决方案的新颖性以及缺乏类似方案——请指出我们可能遗漏的任何内容。）
查看原文
The two of us (PhD in Physics and Mathematics and Philosophical Anthropology) have spent many years working on a mathematically protected secret voting system suitable for making socially significant decisions. During the design process, we realized that such a system must be a peer-to-peer network satisfying six absolute requirements.
We have not been able to find a single existing protocol or academic paper that simultaneously satisfies more than two of them:
First: anonymous user authentication that guarantees the uniqueness of the user's representation in the network (preventing one person from creating multiple accounts) without storing their personal data and without using any external resources, such as certificate authorities or a trusted third party.
Second: a network providing such anonymous authentication must be capable of covering all inhabitants of the Earth. Currently that is just over 8 billion people, but this number grows by more than 100 million every year. Moreover, the deceased must not be excluded from the system, which means monotonic growth is endless.
Third: a network that guarantees anonymity must not have any distinguished nodes—that is, it must be strictly peer-to-peer, consisting of nodes formed by users' devices. However, each user may have not one but several devices, and devices can fail, get lost, become obsolete, or even simply go out of fashion. This immediately changes the scale of the network—it must handle hundreds of billions of nodes.
Fourth: it is necessary to provide each user with access to their account from any device without using passwords, tokens, seed phrases, or anything similar.
Fifth: the key requirement becomes performing a search that enables user recognition in less than 1 second (so that it is imperceptible to the user). That is, any device must locate any other device or data in less than 1 second, while the amount of routing information stored by a single device must be negligibly small, preferably under 100 KB.
Sixth: to guarantee data preservation on unreliable user devices, first, the replication factor must be ~100 or even ~1000x, and second, there must be verification and recovery of routing and data based on consensus without a central arbiter or trusted nodes. At the same time, background synchronization traffic must remain below a few hundred bytes/sec per device to avoid overwhelming mobile data plans.
Our question to HN:
Is the community aware of any existing P2P protocol, DHT, or academic proposal that even attempts to simultaneously satisfy these constraints? We have studied Kademlia, Chord, BATMAN, GNUnet, I2P, etc., and all of them seem to fail at scalability or traffic efficiency. Are we missing something fundamental, or is this truly a blind spot?
(This question is motivated by the fact that we have developed a candidate architecture that, in our opinion, solves all six problems. A brief description can be found on our GitHub [https://github.com/ikhrabry-spec/A-P2P-Network-Architecture-for-100B-Nodes]. Patent applications have been filed, and one patent has been granted. However, we doubt the novelty of our solutions and the absence of analogues—please point us to anything we may have missed.)