Yes you might be missing some context. Let me take it from the beginning.
11 Oct 2022, 20:18
Yes you might be missing some context. Let me take it from the beginning.
A shard is a general term (from database terminology). It simply means storing/processing data in multiple parts when you have lots of data to process. A shard is usually associated with a computer.
When you have lots of data to process, you reach for sharding as a way to organize your data. Shards are unrelated to the size of validator nodes or number of miners in a p2p network. Sharding is a data partitioning concept.
If you look at the Coinweb logo, you can imagine each dot being a Coinweb shard associated with / sitting on top of a corresponding L1 blockchain.
Note: No data enters the Coinweb system except through being written to an L1. This is what makes it even possible for Coinweb to be deterministic. An L1 system is always non-deterministic because choosing which data to put into an L1 block is a fundamentally non-deterministic operation.
Communication can happen between (Coinweb) shards by sending data from other connected shards (like the lines in the Coinweb logo). In Coinweb, not all shards can communicate *directly* with other shards, but indirect communication is possible.
Which shards are connected to which shards and the properties for how they can communicate is called "the delay graph". One important property when information is sent from one shard to another is *the delay*.
When shard A sends a message to shard B, the message is not visible to shard B in the next block in shard B. Instead, the message has to wait for some time before it is visible to shard B. This is because the shards are (in this example) subject to different consensus systems. One might be on top of Ethereum and one might be on top of BNB Chain. If one of those chains had a reorganization, it would cause a reorganization of connected shards without such a delay.
The assumptions in Coinweb is that 1) the probability of a reorganization of blockchain A of size 10 is much less likely than a reorganization of size 1. Further, 2) that no matter how stable one blockchain is (A), and how unstable another blockchain is (B), it is possible to choose a reorganization size for B that is much less likely than a small reorganization of blockchain A. Example assumption: if the probability of a 5 block reorganization of the Bitcoin chain is X, then a 10.000 block reorganization of BNB Chain is much smaller than X.
This is true even if Bitcoin is a "stronger" blockchain than BNB Chain.
Assumptions around these probabilities drive the delays in the delay graph. The delay graph should, in the normal case, make reorganizations of blockchain A, and thus the Coinweb shard associated with blockchain A, not visible to its neighbors.
(Note that Coinweb cross-chain transactions are still consistent even if this assumption is broken)