Skip to content

Latest commit

 

History

History
49 lines (47 loc) · 3.78 KB

TODO.md

File metadata and controls

49 lines (47 loc) · 3.78 KB
Raw

  • udp.c:

    • Could get rid of 'smol' in here, no other processing is performed in this thread

  • TLVList:

    • The 'Pointer' could be directly used in the TLVListIterator, makes it common
    • Not too happy with the way iterator_consumer is done for ContainerIterator, we could just zip the internal ListIterator instead?
    • Implement the IntoIterator Trait as well for the TLVElement. This was done earlier, but I backtracker after I ran into same lifetime issues

  • Some configurable values like number of exchanges per session, number of sessions supported etc, can be bubbled up to some configurator for this crate. I wonder how that is done.

  • About outgoing counter, is it incremented if we send mutliple acknowledgements to the same retransmitted packet? So let's say peer retransmits a packet with ctr 4, for 3 times. Our response ctr, is, say 20. Then should we respond with 20, 21, 22, or 20, 20, 20?

  • I had to use Box::new() to pin ownership for certain objects. Not yet able to use try_new() in the stable releases, and I am not a fan of APIs that panic. We should mostly look at things like heapless:pool or stuff. These objects should really be in the bss, with a single ownership.

  • It might be more efficient to avoid using .find_element() on TLVs. Earlier it was created this way because the spec mentions that the order may change, but it appears that this is unlikely, looking at the C++ implementation. If so, we could be faster, by just specifying looking for tag followed by value.

  • PASE:

    • Pick some sensible and strong values for PBKDF2{iterCnt and Salt-length} based on SoC capability
    • Verifier should only store w0 and L, w1 shouldn't even be stored
    • Allow some way to open the PASE window
    • Allow some way to pass in the 'passcode' and 'salt'
    • In case of error in any of the legs, return StatusReport
    • Provide a way to delete the exchange
    • SPAKE2+: the check with I (abort if h*X == I), as indicated by the RFC is pending

  • Implement the ARM Fail Safe and Regulatory Config properly. Currently we just ack them to proceed further

  • Currently AEAD, sha256 etc are directly used from rust crates. Instead use implementations from openssl/mbedtls - Done. Upstream MRs pending

  • rust-mbedTLS: We have to do some gymnastics because current APIs only support signature encoded in ASN1 format. Fix this upstream

  • CASE:

    • Handle initial MRP Parameters struct from Sigma1

  • FailSafe:

    • Enable timer and expiration handling for fail-safe context

  • Cert Verification:

    • Time validation (Not Before/Not After)
    • KeyUsage flags and others are pending

  • Transport Mgr:

    • Add plain_encode and proto_encode in Packet
    • A new proto_tx should be created in the acks_to_send loop also, otherwise, there is a potential chance of reuse
    • 'transport' object's ownership needs to be inside session, or in the least 'exchange'
    • Sending 'close session' is pending on session reclamation because 'transport' object isn't owned
    • Convert the SessionHandle to &Session? Why maintain a separate object for this?

  • Exchange:

    • What should happen when an exchange is closed by the higher layer, our tx-retrans is pending, and we got a retrans for that exchange?

  • ACL:

    • Device-Type based ACLs
    • NOC CAT
    • Applying ACLs to commands (requires some restructuring of the commands)
    • I think we can the encoder to AccessReq Object making it a complete object for access within the DM
    • List processing of attribute write is missing in IM. List behaviour is add/edit/delete. Currently we only do 'add'

  • Interaction Model

    • List processing of write attributes is different (delete, modify, edit), needs to be handled

  • DataModel:

    • Shall we use a CmdEncoder as a parameter for all the handle_commands()?
    • Need to define common data types for cluster_id_t, endpoint_id_t so their sizes are constantly defined somewhere