# Analysis of bumping `eth` protocol version > This document is the successor to [this document](https://notes.ethereum.org/M41Oj3WyRVSC36Rr0c-Tjg) Goal of this document is to outline the client behaviors and consequences related to bumping or NOT bumping the `eth` protocol version with respect to 4444s and serving pre-merge data. ## Cases to Consider We will try to enumerate the following cases. - Clients that do not update and remain on old client versions - Clients that do update - Clients that drop the pre-merge data - Clients that retain the pre-merge data - Clients that are "full" syncing history and need to obtain full history. ### Case 1: Interactions between updated and not-updated clients #### With Bumped version Clients that do NOT update will remain on an old version of the client software that is not aware of any of the new protocol changes. Clients that do update *can* retain support for both the old and the new protocol version assuming they also retain the pre-merge data. Clients that drop the pre-merge data can only support the new protocol version and would not have to deal with requests from not-updated clients. Clients can implement the new protocol version without dropping the underlying data. This in theory allows for a simpler and more gradual implementation path. Clients that implement the new protocol version will still be required to implement handling for out-of-bound data requests on the new protocol version, likely by disconnecting or returning and empty response. Clients that do NOT update will face a diminished group of peers which support their outdated protocol version. Discovery of peers that continue to support the old protocol version may be difficult. #### Without bumped version They will continue to try and request pre-merge history during their sync process using the latest protocol version they are aware of. These requests will succeed when connected to another not updated client or an updated client that chooses to both retain the pre-merge data **and** to serve it. Clients that do update will continue to receive requests for pre-merge data from not-updated clients and will have to decide how to handle these requests. Obvious choices for handling include disconnection and an empty response. ### Case 2: Dropping The Data #### With bumped version A client that chooses to drop the pre-merge data will no longer be capable of serving this data. A client can clearly signal that they will not serve pre-merge data by only advertising support for the new bumped protocol version. A client that chooses to keep the pre-merge data could continue to support the old protocol version as long as they wish. #### Without bumped version A client that chooses to drop the pre-merge data will be indistinguishable from a client that still serves the data on the network until requests are made for that data. The protocol will be populated with a mix of clients that both still have the pre-merge data and clients that do not. Some of the clients will continue requesting this data and some will not. Clients will have to implement handling for pre-merge data requests depending on whether they have this data or not. Some clients may continue to implement full sync over this protocol as it will still work given a sufficient number of clients on the network are still responding to data requests. Clients that choose to keep the pre-merge data could continue to serve it. ### Case 3: Full Syncing #### With bumped version A client that does "full" syncing will be able to know that they cannot request this data over the new bumped protocol. Full syncing clients will need to implement either a separate protocol or a non-standard use of the bumped protocol. Clients that choose to drop the data would not receive requests for pre-merge data. #### Without bumped version A client that does "full" syncing could make no changes. A client that does "full" syncing would experience performance degredation to its sync process if a sufficient number of the nodes on the network stop serving pre-merge data. Clients that choose to drop the data would continue to receive requests for pre-merge data from "full" syncing clients that choose to continue to use the old protocol behavior. ## Summary No matter what choice we make, clients will need to implement logic for handling requests for pre-merge data. Negatives effecting clients that choose not to update seem like they are inconsequential and probably not worth spending time on. There is no requirement that the protocol version be bumped. The most compelling *meta* reason for bumping seems to be as a sort of "forcing function". The bump requires clients to act. Not bumping allows for innaction. Bumping seems to make for a cleaner separation between those that act and those that don't act, with small benefits to those that updated, and larger consequences for those that do not.