HttpParser and multiple concatenated HTTP responses

[mejedi]

Hi!

If my understanding is correct HttpParser.execute returns the number of bytes consumed, hence it is possible to parse multiple concatenated HTTP responses by checking is_message_complete() and starting a new parser from the same very position the last one stopped.

However the observed behavior is very different: execute() happily consumes the whole string, the subsequent responses text ends up in the message body and _clen_rest is negative.

[benoitc]

Do you mean in case of pipelined requests?

[mejedi]

Precisely.

Actually I am working on a weird task motivated by the needs of a network analysis tool. I am looking for HTTP message boundaries in opaque byte streams.

Currently I ended up using the modified parser from httplib + eventlet so that the script can do other things while waiting for more data to arrive.