The latest released version is supported. Fixes go into the next release rather than into patches of older ones.

Report privately through GitHub: open a security advisory. That reaches the maintainers without the report being public, and is the only channel to use for something exploitable. Please do not open a public issue for it.

Expect an acknowledgement within a week. What happens next depends on what it is: a fix with a released version, an explanation of why it is not a vulnerability, or a question if the report needs more to reproduce. You will be credited in the advisory unless you would rather not be.

Anything that is not a vulnerability — a protocol misread, a callback that fires at the wrong time, a negotiation that stalls — belongs in a public issue, where it is easier to discuss.

This library reads bytes from a peer that is, by construction, not trusted: a stranger connecting to your server, or a server your client dialled. Everything it parses is attacker-controlled. Reports worth making privately include:

• Memory exhaustion. An input that makes the interpreter accumulate without bound. Every accumulator is supposed to have a ceiling — see Limits on untrusted input — so an unbounded one is a bug in this library, not a tuning question.
• Escape from the negotiation stream. Protocol bytes reaching OnSubmit as if a user had typed them, or user input being read as protocol. Both sides of that boundary matter: an MCP message that a player can forge by typing, or a subnegotiation whose payload is delivered as a line.
• A parser that can be made to crash or hang the byte-processing loop, rather than dropping the message and carrying on.
• Anything that leaks the host. This library sends no environment variable, user name or terminal it was not explicitly given — see ENVIRON and NEW-ENVIRON. Something that does is a report.
• A supply-chain problem in what is published: the package contents, the release workflow, or the provenance attestation.

• Telnet is a cleartext protocol. RFC 854 has no confidentiality and no integrity; anything on the path can read and rewrite the session. That is the protocol, not a defect in this implementation. Put TLS underneath it if that matters.
• AuthenticationProtocol (RFC 2941) and EncryptionProtocol (RFC 2946) are negotiation frameworks, not cryptography. This library carries the messages; your callbacks implement the mechanism. RFC 2946's own algorithms — DES, 3DES, CAST — are long broken, and nothing here implements them. A report that "the DES mode is weak" is a fact about RFC 2946.
• MCCP's expansion ceiling is a policy, not a proof. A peer's compressed stream may expand 200:1 by default (.WithMaxExpansionRatio), on top of the downstream line-buffer and subnegotiation limits. A peer that stays under the ceiling and still makes this side do disproportionate work is a report; a peer that is refused at it is the feature working.

There are no long-lived publishing keys. Releases are built and published by the release workflow from a tag whose commit is an ancestor of main, pushed to nuget.org through trusted publishing, and carry an attestation of the workflow and commit that built them.

Verify a downloaded package with the GitHub CLI:

gh attestation verify TelnetNegotiationCore.3.0.0.nupkg --repo HarryCordewener/TelnetNegotiationCore