A malicious or compromised server can send a reply line that never ends, which the client buffers until memory runs out. A reply made of many short lines also keeps a CPU core busy: for a few seconds per reply since 0.11.22, for as long as the server wants before it.
Applications that connect only to a trusted server over implicit TLS are not
exposed; the risk applies to clients that may connect to untrusted servers, or
that use STARTTLS or no TLS, where a man-in-the-middle can forge the reply.
{
"license": "CC0-1.0"
}