Let's assume there are some unforeseen challenges in the future that require the IOTA network to be adapted. I've read there is no hard-forking possible in IOTA. Hence, would it be problematic to push protocol changes at a future date?
For example if there is an important case for including fees on every transaction (this is just a hypothetical scenario!)—how can this be achieved? It would require all the nodes to be updated manually?
Currently, on the one side all Full Nodes run the IOTA Reference Implementation in order to maintain the Tangle in an distributed manner, and on the other side the official IOTA wallet uses these Full Nodes to access the Tangle for receiving and sending transactions. In future, we might see additional implementations providing the same APIs.
The Full Nodes are the entry points into the IOTA network. They are providing interfaces for reading and extending the Tangle; their most important tasks are:
If someone wants to introduce some protocol adaption, this has to be implemented in the software that is running on the Full Nodes. For examlpe, a Full Node may decide to run a software that requires the payment of a certain transaction fee for each transaction in order to process it. Then, the Full Node's behavior must be accepted by multiple parties. Namely, two requirements have to be satisfied:
If the first requirement is not met, the Full Node will not be able to serve IOTA users, which would make the Full Node obsolete. If the second requirement is not met, the Full Node will not be able to get a full view on the Tangle as information about transactions from other Full Nodes are missing, which in the end would make the Full Node obsolete, too.
Hence, future changes to the IOTA protocol are possible if they are adapted by "enough" IOTA users and "enough" Full Nodes. However, coming up with concrete numbers is a very difficult task. Statements about the required share of IOTA users and Full Nodes would require comprehensive simulations that rely on (a) knowledge of the Full Node infrastructure, and (b) the occurring protocol adaption behavior of both IOTA users and Full Node operators. As the detailed full node infrastructure by design is not fully known, and as there will always be a high uncertainty in the adaption behavior, only very rough estimates can be made (at some point a consensus may be reached, at some point a certain attack may be possible, at some other point some other attack may be possible, etc.).
In general, protocol adaptions can be separated into two major categories:
A soft fork change would still allow all IOTA Full Nodes, independent of whether they are upgraded or not, to maintain a collective consensus on the state of transactions in the Tangle.
A hard fork change would result in two options:
In Bitcoin, we could observe hard forks already multiple times. As a result, in addition to the traditional Bitcoin we now have further Bitcoin blockchains/coins such as Bitcoin Cash (larger block size) or Bitcoin Gold (different hashing algorithm). For realizing protocol adaptions, Bitcoin usually uses polls where miners can indicate their adaption intention prior to a planned adaption via a flag in the header of their mined blocks (i.e. the miner community agrees to vote on whether a discussed protocol adaption should be realized or not). Based on the poll results, the miners can decide how they want to proceed.
For IOTA, the rather anonymous nature of the Full Node infrastructure seems to make it harder to agree on protocol changes in such a community-based manner.
Nodes will have to update to the new IRI node software implementation that is released if there is some desire to change the protocol.
As nodes are connected to neighbors, node operators who update may disconnect from un-updating neighbors. In this way, consensus needs to be reached among the network of node operators on what implementation of IRI to run.
As nodes are run by by users and require maintenance unlike transactors which may be computers and IoT sensors, it makes sense that they will be updatable.
So far, protocol updates happen with almost every new iri release. But IOTA is still in beta (and so far the changes were uncontroversal, so nodes updated quickly).
So, yes, protocol will probably be adapted a lot in the (near) future :-)
At some point in the future, I think all changes should be as backward compatible as possible, but probably there will still be changes.
Also, about hard forks, see this question.
