It's much more convenient to decode bytestreams (e.g. for a wire protocol) using % formatting. The lack of % formatting for bytes was a huge impediment for porting mercurial to python3, just as an example.
Could you point me to a code example? It's hard to see how old-style formatting would facilitate decoding. Typically, I think of decoding as being a process which deconstructs something; not like formatting, which is really more about constructing.
I remember Armin Ronacher (author of many nice things, especially Flask) even wrote a blog post against Python 3 and this was one of things he hated about it.
