First check out the Help Wanted tag in the issues section for specific ideas.
Right now the encoder / decoder works by encoding or decoding a full buffer. SLIP (and COBS?) encoding, for instance, can be encoded / decoded on the fly, allowing for a smaller buffer allocation. https://github.com/CNMAT/OSC uses an "on-the-fly" approach, rather than a large buffer approach. It would be interesting to investigate this.