Parsing HTTP

Previously...

• I decided to keep using the ejabberd BOSH server.
• I started messing around with nodejs TCP servers.
• I switched ECM over to using a straight nodejs TCP server.

Once ECM started sending straight HTTP, I needed to change megaphone to be able to parse it.  I was planning on using erlang:decode_packet to do this, but I was running into a problem.  The input looked like this:

POST /http-bind/ HTTP/1.1

Host: ubuntu2

User-Agent: Pidgin 2.6.6 (libpurple 2.6.6)

Content-Encoding: text/xml; charset=utf-8

Content-Length: 225

...

The first call to decode_packet returned

{http_request,'POST', {abs_path,"/http-bind/"}, {1,1}}

With the rest being the stuff that started at "Host:"  Note that each line is separated with a carriage-return/linefeed combination.  

Thank you DOS for leaving us with that legacy, but I digress.

The second call with the the binary start therefore starts with the second line was returning:

{http_error,"Host: ubuntu2\r\n"}

which was not exactly what I expected.  After a bit of puttering about I discovered the following from the erlang documentation for decode_packet:

The protocol type http should only be used for the first line when a HttpRequest or a HttpResponse is expected. The following calls should use httph to get HttpHeader's until http_eoh is returned that marks the end of the headers and the beginning of any following message body.  

Of course I knew this all along and I was just testing the readers (OK, I guess it's just singular) to see if they would catch this point.  

So I was testing you.  Yeah.  It was not that I myself needed to RTFM.

At any rate, the next step is to get megaphone_socket to maintain some state in between calls so that it will keep the data from call to call and use the correct data for each invocation of decode_packet.