We may need both actually: solid protocol and great reference implementation.
Wayland was advertised as protocol mostly. After almost 20 years, it still has not promised to be a full replacement to xorg and probably never will, where wayland developers say "this is not our goal". Took people quite some time to realise that; it's been more recently become obvious, but say, 12 years ago few understood this.
I still have not been able to find working replacements to all that works on xorg, for instance; specifically imagemagick is different on wayland. I may try again at a later point in time, but my old workings there did not work, and replacements seem dead or ineffective or incomplete - that is quite frustrating for something that was aggressively advertised as "this is now the future".
100% agree, I should have phrased it less black and white. In theory a spec is all you need but in practice solid reference implementations can iron out ambiguity plus it can uncover issues in the spec.
Wayland was advertised as protocol mostly. After almost 20 years, it still has not promised to be a full replacement to xorg and probably never will, where wayland developers say "this is not our goal". Took people quite some time to realise that; it's been more recently become obvious, but say, 12 years ago few understood this.
I still have not been able to find working replacements to all that works on xorg, for instance; specifically imagemagick is different on wayland. I may try again at a later point in time, but my old workings there did not work, and replacements seem dead or ineffective or incomplete - that is quite frustrating for something that was aggressively advertised as "this is now the future".