OK, so maybe the fact that I am trying to analyze what segwit softfork in the upcoming weeks will do, that explains my not understanding that future upgrades with a new signature scheme are part of the analysis... Would these changes require a hardfork, or the usual softfork can change the signature scheme? It is kind of hard to analyze something based on unspecified future upgrades with a different signature scheme.
Well I think those changes could be soft forked in because it changes the script version number, which I think would only affect the address type. I could be wrong though.
maybe there can be just a single aggregated signature for all the tx in a block? I have no idea if that is possible, but if it is, then that could be added to the coinbase and then we wont need any witness data at all. Did I get that right?
I am fairly certain that this isn't possible since it would require the private keys that can spend the inputs of all of the transactions to sign it. However, I could be wrong as I am not well versed in many parts of cryptography. There maybe is an algorithm which could combine all of the signatures, I don't know. You'll have to ask gmaxwell, he is the "chief cryptographer".
I would think that to implement a blockwide aggregated signature, would at the least require a three step process:
1. block is mined to determine the tx that are in it
2. the txids of this protoblock would need to be broadcast
3. nodes that are running and part of the protoblock txid would need to sign and return to miner(s)?
4. miner prunes out all the signatures that are aggregated and publishes optimized block
Not sure if the libsecp256k1-zkp lib's schnorr routines are sufficient for this and clearly it cant be done with all sigs, and of course details about timing and protocol for the above have plenty to be defined. like when is the mining reward earned, etc. so this is just a fantasy protocol for now
I am not saying the above is possible, just that the above is the minimum back and forth that would be needed and it has some privacy issues, so some privacy enhancements are probably needed too. A bitmap of the aggregate signers would probably be needed, but that can be run length encoded to take up relatively small amount of space