
Within the last Mempool article, I went over the totally different sorts of relay coverage filters, why they exist, and the incentives that in the end resolve how efficient every class of filter is at stopping the affirmation of various courses of transactions. On this piece I’ll be trying on the dynamics of the relay community when some nodes on the community are working totally different relay insurance policies in comparison with different nodes.
All else being equal, when nodes on the community are working homogenous relay insurance policies of their mempools, all transactions ought to propagate throughout the complete community provided that they pay the minimal feerate needed to not be evicted from a node’s mempool throughout occasions of huge transaction backlogs. This adjustments when totally different nodes on the community are working heterogenous insurance policies.
The Bitcoin relay community operates on a finest effort foundation, utilizing what is known as a flood-fill structure. Which means that when a transaction is obtained by one node, it’s forwarded to each different node it’s linked to besides the one which it obtained the transaction from. This can be a extremely inefficient community structure, however within the context of a decentralized system it gives a excessive diploma of assure that the transaction will ultimately attain its supposed vacation spot, the miners.
Introducing filters in a node’s relay coverage to limit the relaying of in any other case legitimate transactions in idea introduces friction to the propagation of that transaction, and degrades the reliability of the community’s means to carry out this perform. In follow, issues aren’t that easy.
How A lot Friction Prevents Propagation
Let’s have a look at a simplified instance of various community node compositions. Within the following graphics blue nodes signify ones that will propagate some arbitrary class of consensus legitimate transactions, and purple nodes signify ones that may not propagate these transactions. The collective set of miners is denoted within the heart as a easy illustration of the place transacting customers in the end need their transactions to wind up in order to ultimately be confirmed within the blockchain.

This can be a mannequin of the community during which the nodes refusing to propagate these transactions are a transparent minority. As you may clearly see, any node on the community that accepts them has a transparent path to relay them to the miners. The 2 nodes making an attempt to limit the transactions propagation throughout the community don’t have any impact on their eventual receipt by miners’ nodes.

On this diagram, you may see that just about half of the instance community is instituting filtering insurance policies for this class of transactions. Regardless of this, solely a part of the community that propagates these transactions is lower off from a path to miners. The remainder of the nodes not filtering nonetheless have a transparent path to miners. This has launched some extent of friction for a subset of customers, however the others can nonetheless freely have interaction in propagating these transactions.
Even for the customers which are affected by filtering nodes, solely a single connection to the remainder of the community nodes that aren’t lower off from miners (or a direct connection to a miner) is critical to ensure that that friction to be eliminated. If the true relay community have been to have an identical composition to this instance, all it might take is a single new connection to alleviate the issue.

On this state of affairs, solely a tiny minority of the community is definitely propagating these transactions. The remainder of the community is partaking in filtering insurance policies to stop their propagation. Even on this case nevertheless, these nodes that aren’t filtering nonetheless have a transparent path to propagate them to miners.
Solely this tiny minority of non-filtering nodes is critical in an effort to guarantee their eventual propagation to miners. Preferential peering logic, i.e. performance to make sure that your node prefers friends who implement the identical software program model or relay insurance policies. All these options can assure that friends who will propagate one thing to others gained’t discover one another and preserve connections amongst themselves throughout the community.
The Tolerant Minority
As you may see taking a look at these totally different examples, even within the face of an amazing majority of the general public community partaking in filtering of a particular class of transactions, all that’s needed for them to efficiently propagate throughout the community to miners is a small minority of the community to propagate and relay them.
These nodes will basically, by no matter technical mechanism, create a “sub-network” inside the bigger public relay community in an effort to assure that there are viable paths from customers partaking in some of these transactions to the miners keen to incorporate them of their blocks.
There may be basically nothing that may be executed to counter this dynamic besides to have interaction in a sybil assault in opposition to all of those nodes, and sybil assaults solely want a single trustworthy connection in an effort to be fully defeated. As nicely, an trustworthy node creating a really massive variety of connections with different nodes on the community can increase the price of such a sybil assault exorbitantly. The extra connections it creates, the extra sybil nodes have to be spun up in an effort to eat all of its connection slots.
What If There Is No Minority?
So what if there is no such thing as a Tolerant Minority? What’s going to occur to this class of transactions in that case?

If customers nonetheless need to make them and pay charges to miners for them, they are going to be confirmed. Miners will merely arrange an API. The position of miners is to substantiate transactions, and the rationale they achieve this is to maximise revenue. Miners should not selfless entities, or morally or ideologically motivated, they’re a enterprise. They exist to earn a living.
If customers exist which are keen to pay them cash for a sure sort of transaction, and the whole lot of the general public relay community is refusing to propagate these transactions to miners in an effort to embody them in blocks, miners will create one other means for customers to submit these transactions to them.
It’s merely the rational transfer to make as a revenue motivated actor when prospects exist that want to pay you cash.
Relay Coverage Is Not A Alternative For Consensus
On the finish of the day, relay coverage can’t efficiently censor transactions if they’re consensus legitimate, customers are keen to pay for them, and miners should not have some extenuating circumstances to show down the charges customers are keen to pay (resembling inflicting materials harm or hurt to nodes on the community, i.e. crashing nodes, propagating blocks that take hours to confirm on a shopper PC, and so on.).
If some class of transactions is actually seen as undesirable by Bitcoin customers and node operators, there is no such thing as a answer to stopping them from being confirmed within the blockchain wanting enacting a consensus change to make them invalid.
If it have been potential to easily stop transactions from being confirmed by filtering insurance policies carried out on the relay community, then Bitcoin wouldn’t be censorship resistant.