Should/can we pass around strings rather than soup objects?

[nickmoreton]

The first time we get some soup is with the lxml parser.

Are you saying that the soups are different because one uses lxml and another uses html.parser? If so, then that's a strong argument for passing around string HTML objects instead of soup instances.

I favour readability and consistency at this stage, especially as we aren't yet aware of any performance concern.

Originally posted by @nimasmi in #57 (comment)