Wait wait wait, what do you mean XMPP over TLS?

Okay so there is a high chance that you don't know what XMPP is, unless you like reading RFCs or you have had to implement some sort of chatting feature recently.

XMPP is a protocol based on XML, to send and receive messages, as simple as that.

Streams

XMPP is just a set of simple rules for defining a XML stream yeah, you heard me a stream. If you haven't found any XML streams yet, it's basically like a XML document but for things that require to be networked.

Wait but I have never seen xml:// in my browser how do those streams, stream.

Well for example XMPP supports:

• WebSocket
• QUIC
• TCP
• TCP/TLS

That's all nice and good till you realize, that's raw tcp, there is no framing.

Wait wait wait a moment here, what the fuck is framing and why are you dying out here for it?

Framing

If you know a bit about how strings work in C you might know that they are something like this:

Where you have your data H E Y and then you have a special character \0, that character is a null terminator.

The computer would do:

Okay but, like, why you telling me about C you fuc-

When dealing with networking we also need a way, to tell the computer to stop reading.

Apart of including a marker like C does with it's null terminator (\0). We can also tell the computer the number of bytes to read, like in HTTP:

This are two ways of telling the computer to stop reading, either from memory or network, so that it doesn't continue forever.

The problem

Let me talk you little sh-

We just established, that XMPP has several protocols, and weirdly enough what framing is, so whats the deal?

Neither XMPP/TCP or XMPP/TCP/TLS have FRAMING.

Wait didn't you just say that meant, reading forever?

About that.... Do you remember that XMPP was just an XML stream, right? A lot of XML are parsers are stream parsers, meaning that they just go reading the XML and do events like start-tag, start-attribute, end-tag, meaning that they don't need the whole document for parsing, it just processes forever.

Yet, my problem is that I need to get the whole "frame".

That's a bit selfish of you not gonna lie

The solution

My technical solution went as follows:

Best case

• Pull for data from the peer
• Check if XML is a valid fragment
• Data is valid
• Process it

Large fragment

• Pull for data from the peer
• Check if XML is a valid fragment
• Data is invalid
• Push to a buffer
• Pull for data from the peer
• Push data to buffer
• Check if buffer is valid XML
• Data is valid
• Process it

Was it that hard?

To get the buffering, async reading from two streams at once, processing data in time, returning the right data, was very tedious. But now I learnt my lesson not to fall for this pitfalls if I ever make a protocol.

Conclusion

Please add framing to your protocols, and please please don't try to buffer the protocols that don't have framing.