Explore if nbd_aio_notify_error is needed for faster response if
server goes away.

Bindings in other languages.
 - Latest attempt at adding Rust:
   https://www.redhat.com/archives/libguestfs/2019-August/msg00416.html

Example code integrating with ppoll, pollfd, APR pollset (and others?).

Example command line utils to copy in/out (like qemu-img convert).

NBD resize extension.

TLS should properly shut down the session (calling gnutls_bye).

Performance: Chart it over various buffer sizes and threads, as that
  should make it easier to identify systematic issues.

Examine other fuzzers: https://gitlab.com/akihe/radamsa

nbdcopy:
 - Use parallel AIO operations if the input or output is a regular
   file or block device.
 - Consider using more threads.
 - Create sparseness in output.
 - Handle extents efficiently.
 - Write zeroes efficiently.
 - Allow copy between NBD servers.

nbdfuse:
 - If you write beyond the end of the virtual file, it returns EIO.
 - Implement trim/discard.
 - Implement write_zeroes.
 - Implement block_status.
 - Could be made multithreaded for improved performance.

nbdinfo:
 - info-list.sh and info-list-json.sh tests should use nbdkit + --run
   when nbdkit supports listing exports.

Suggested API improvements:
  general:
  - synchronous APIs that have a timeout or can be cancelled

  connecting:
  - nbd_connect_tcp: allow control over whether IPv4 or IPv6 is desired
  - nbd_connect_uri: allow control over which features are enabled
  - nbd_connect_command: allow passing char **env
  - connection completed callback

  TLS:
  - should be individual APIs for setting each TLS file
    (set_tls_certificates can continue to exist)
  - TLS should have a way to pass in PIN or password to
    support encrypted keys
  - allow use of TLS memory APIs (gnutls_datum_t + gnutls_*_mem)
  - support PKCS11 URIs (RFC 7512)

  Easier server implementation testing:
  - a way to force NBD_OPT_EXPORT_NAME over NBD_OPT_GO
  - a way to forcefully violate protocol (such as allowing writes to a
    readonly connection, or sending a request that exceeds bounds) for
    testing server reactions

  subprocess
  - it should be possible to use nbd_close and never block, so
    maybe nbd_shutdown should wait for the subprocess or there
    should be another API to do this
  - capture error message when nbd_connect_command fails