Deterministic addressing for serum markets

[armaniferrante]

It would be great to have a deterministic address for a market, given the base and quote currency. We can do this by having market accounts initialized in a program derived address, using an oracle for lot size configuration. One problem that needs to be considered is how to deal with the case where a price changes enough that the existing lot size configuration no longer makes sense.

[tomjohn1028]

using an oracle for lot size configuration

@armaniferrante can you fill me in on the above? When/Why do lot sizes change on a market? How often does this occur?

[armaniferrante]

Lot sizes may need to be adjusted if the value of one of the tokens changes dramatically. I don't have a good answer for how often it occurs.

[cdrappi]

@armaniferrante For PsyOptions use case, we care more about deterministic addressing on (baseToken, quoteToken, baseLotSize, quoteLotSize) rather than (baseToken, quoteToken) which might simplify this problem a lot. If so, is this potentially as simple as changing the market initialisation from generating a random address to one that creates one as a function of those 4 params?

[armaniferrante]

Related. Once we have "permissioned markets", it would be nice to have a governable market list for each permissioning entity on chain.

[ryoqun]

hi eng from @solana-labs. :)

i think the problem is that downstream programs need to hard-code market id?

fyi, I'm trying to make specifying the market id in the tx needless for the downstream programs to begin with the aid of rpc tx sim:
solana-labs/solana#17796

so, this proposal eventually enables invoked program to dynamically load accounts at arbitrary address (by deterministically deriving the address from some other important parameters like baseToken, quoteToken).

Please have a look and leave comments if any. I'm especially wondering how's the hard requirement of rpc sim could affect latency sensitive defi apps..