The best way out of the definition trap, as far as I can see, is to ignore the entire folk/non-folk dichotomy entirely and just discuss the song.

If we react to the mass of evidence so well researched and presented by Steve Roud by closing down the debate, then his time will have been wasted. The whole point of this book is to open up the question of what is 'folk'. Roud himself describes that question in his 'Afterword' (yes, I've been dipping again) as 'the elephant in the room'.

What I (and I suspect a lot of us on this thread) have always understood as 'traditional folk song' has been based broadly on the concept as erected by Victorian / Edwardian collectors. Roud has compiled evidence that a wide range of additional songs were on the lips of the working classes of the day. If Sharp et al were justified in rejecting contemporary pop songs, then the edifice still stands. If not, then the body of material labelled 'folk song' is - not 'fake', certainly - but an unrepresentative sample. That's a bigger philosophical question than whether Steve G's farmer's original compositions should be called 'folk' or not. Without addressing it, how could one even attempt to compile a collection of 'English Folk Songs' when a publisher like, say, Penguin Books, came calling?

Roud's concluding sentence affirms his view that traditional process is of prime importance in his view of this music. With that, Cecil Sharp would agree. But don't let's all throw up our hands and cry "Oh no, another Mudcat 'what is folk?' food fight!" when this is a distinctly different debate from the one about Dylan / Mumfords etc etc.