Wenjing Chu

We went over the spreadsheet that Wenjing has prepared to quickly resolve issues (see screenshot #1). We started with HOP codes (see screenshot #2).

Sam Smith explained why the CESR design supports pipelining using the XHOP payload structure.

We discussed how this design optimizes the ability for a core processor to process as many messages as possible because it allows the pipe to shunt off messages extremely efficiently to different cores.

We discussed the ThreadID field at the tunnel level. Wenjing explained that some messages have nonces. Do we need both? Sam suggested that it should be "up-layered". Wenjing asked if it can be efficiently performed by an up-layer. Wenjing was convinced to not include a ThreadID at the base TSP layer because: a) not every up-layer protocol needs threading, and b) if they do need threading, they may need different kinds of thread IDs.

We next discussed the relationship formation protocol. The question there was whether we needed a nonce. Sam pointed out that the relationship formation protocol requires VIDs, and VIDs need to be globally unique, so a separate nonce should not be needed.

We next discussed codes for cyphersuites. Wenjing asked about the HPKE algorithms. Sam said they could be added to the primitive code table. See screenshot #3. Sam explained that CESR code allocation follows the Pareto Principle: single-character codes are used for the most frequent primitives, then two-character codes that start with 0 are the next set, then all other two-character codes, then four-character codes, then larger codes. It is all designed for efficient parsing.

Next we discussed screenshot #4. Wenjing pointed out that the issue is that HPKE has its own code table. Sam explained that the entire HPKE code table can be prefixed with a single CESR code. Wenjing asked for a pointer to the CESR Master Code table, and Sam gave that pointer to the main CESR spec.

Next we discussed signatures—see screenshot #5. That was a short discussion.

Wenjing