abread / sec Goto Github PK

Preferrably under some standard like PKCS#12/PKCS#11/PKCS#8, but if that's not possible, just use some key derivation function on the passphrase to compute a symmetric key to {en,de}crypt the private keys.

I think it would be best to define a second struct representing the private part of the keystore on-disk (with all the possibly encrypted shenanigans). This means pretty much everything apart from the keystore open/save calls stays the same.

MAKE SURE THE ENCRYPTION IS OPTIONAL we'll test it once to see if it works and pretty much never use it

Currently crashing mid-server::hdlt_store::HdltLocalStoreInner::save() means a potentially corrupted storage. We should write a new file, and then rename it to the storage path (which is atomic).

The quroum_size concept/method/thing is pervasive in the model and server (the integration tests do mention it, but only because the server demands it). They should move to thinking about max_faults.

Originally posted by @bsdinis in #16 (comment)

pretty much all calls are expensive rn (writes have I/O and block everything else, all reads go through all data)

We could just send an additional value x such that sha256(restofthemessage || hash) = d, where d has the last k bits equal to 0 (the bigger the k, the harder it is for the sender to compute x).

The server itself, with persistence and all that. Since we have no replication or anything like that, we can just plop all LocationProofs in a set/vec and query them as we see fit. Easy stuff.