Waitress implemented a "MAY" part of the RFC7230 (https://tools.ietf.org/html/rfc7230#section-3.5) which states:
Although the line terminator for the start-line and header fields is
the sequence CRLF, a recipient MAY recognize a single LF as a line
terminator and ignore any preceding CR.
Unfortunately if a front-end server does not parse header fields with an LF the same way as it does those with a CRLF it can lead to the front-end and the back-end server parsing the same HTTP message in two different ways. This can lead to a potential for HTTP request smuggling/splitting whereby Waitress may see two requests while the front-end server only sees a single HTTP message.
Example:
Content-Length: 100[CRLF]
X-Header: x[LF]Content-Length: 0[CRLF]
Would get treated by Waitress as if it were:
Content-Length: 100
X-Header: x
Content-Length: 0
This could potentially get used by attackers to split the HTTP request and smuggle a second request in the body of the first.
This issue is fixed in Waitress 1.4.0. This brings a range of changes to harden Waitress against potential HTTP request confusions, and may change the behaviour of Waitress behind non-conformist proxies.
Waitress no longer implements the MAY part of the specification and instead requires that all lines are terminated correctly with CRLF. If any lines are found with a bare CR or LF a 400 Bad Request is sent back to the requesting entity.
The Pylons Project recommends upgrading as soon as possible, while validating that the changes in Waitress don't cause any changes in behavior.
Various reverse proxies may have protections against sending potentially bad HTTP requests to the backend, and or hardening against potential issues like this. If the reverse proxy doesn't use HTTP/1.1 for connecting to the backend issues are also somewhat mitigated, as HTTP pipelining does not exist in HTTP/1.0 and Waitress will close the connection after every single request (unless the Keep Alive header is explicitly sent... so this is not a fool proof security method)