cairn/changes/stream-0-2-migration/proposal.md
io-imap master moved to pimalaya-stream ^0.2, which is unpublished, so himalaya could not take any io-imap update. Patching stream to git did not resolve it: a patch replaces a crate globally with one version, and io-gmail, io-jmap, io-msgraph and io-pim-discovery all required ^0.1, so cargo resolved both 0.1.2 and 0.2.0 into the graph. Two copies of the same crate means two incompatible Tls types, and the wizard stopped compiling where one met the other.
Stream 0.2 is a small release with one breaking change: the sasl module moved out to the io-sasl crate. The std and tls module APIs are byte-identical to 0.1.2, which is what made the four blocked crates cheap to move.
io-imap and io-smtp had each also grown a client trait layer in the same window, so taking their new commits meant absorbing that too.
The four blocked crates take stream 0.2. None of them used pimalaya_stream::sasl, so each is a one-line dependency bump with no code change.
himalaya then absorbs three upstream API changes: pimalaya_stream::sasl becomes io-sasl (with the credential structs renamed Sasl*Creds and SCRAM's nonce and channel binding now carried on the credentials), ImapClientStdConnectOptions becomes ImapSessionOpenOptions, and the io-imap and io-smtp command methods move onto ImapClient and SmtpClient traits that call sites bring into scope. SmtpClientStd::connect takes its starttls flag in an options struct and returns the client with the EHLO capabilities.
No crate versions are bumped and nothing is released. The four crates keep the versions they published under so himalaya's existing version = "0.2" style requirements still match through the patch table.
The wizard keeps offering the same six SASL mechanisms. io-sasl knows sixteen, so the two matches over SaslMechanism gain a catch-all that names the mechanism rather than silently accepting one the config cannot express.
io-oauth also still requires stream ^0.1, but nothing in himalaya's graph pulls it, so it is left alone.