3 ms·

I'd love to see a list of some standards more backwards-compatible than HTML.

Silhouette 15y ago
The problem isn't backward compatibility with previous standards as defined today. That has, as you suggest, generally been excellent. The problem is that if you decide your standard is going to be a "living" document that changes rapidly as the industry evolves, backward compatibility ceases to have any meaning because there is nothing stable to be backward compatible with. What happens when one browser pioneers a new feature, the living HTML5 standard quickly adopts it, developers start writing to that standard, but then two other browsers decide on another way to do the same thing that they think is better and someone changes the living HTML5 standard to reflect the new "standard" instead? One possibility is that the developers' existing code breaks. Another is that browser makers are stuck with supporting both once-standard interpretations for a long time. Alternatively, but almost as bad, what happens when the living HTML5 standard quickly adopts a new feature that is becoming popular, other browsers find a better way to do the same thing shortly afterwards, and the standard doesn't change to match? Now the standard sucks, because it rushed into formalising something before enough consensus had built. Formal standards don't mean anything unless they are robust, stable, citable references that everyone concerned can agree on. A living document that rapidly changes in the face of a fast-evolving industry is exactly zero of those things, and its value will be similarly small and round. It's merely a reference non-implementation, and since we're talking about browsers, we already have several candidates for being an actual reference implementation, just as we did in the old IE vs. Netscape days, when we learned for the first time why formal standards matter. Alas, as George Santayana observed a century ago, those who cannot remember the past are condemned to repeat it.
- yuhong 15y ago
  Well, ESR releases of Firefox are one thing, but browsers do not actually implement different HTML versions separately or as one unit.
- MatthewPhillips 15y ago
  I can't think of any instances where this happened. The closest I can think of is Web SQL, but that was still in a working group when it was canned. Developers who work on in-progress browser features are aware of the risk, or at least should be.

The problem isn't backward compatibility with previous standards as defined today. That has, as you suggest, generally been excellent. The problem is that if you decide your standard is going to be a "living" document that changes rapidly as the industry evolves, backward compatibility ceases to have any meaning because there is nothing stable to be backward compatible with. What happens when one browser pioneers a new feature, the living HTML5 standard quickly adopts it, developers start writing to that standard, but then two other browsers decide on another way to do the same thing that they think is better and someone changes the living HTML5 standard to reflect the new "standard" instead? One possibility is that the developers' existing code breaks. Another is that browser makers are stuck with supporting both once-standard interpretations for a long time. Alternatively, but almost as bad, what happens when the living HTML5 standard quickly adopts a new feature that is becoming popular, other browsers find a better way to do the same thing shortly afterwards, and the standard doesn't change to match? Now the standard sucks, because it rushed into formalising something before enough consensus had built. Formal standards don't mean anything unless they are robust, stable, citable references that everyone concerned can agree on. A living document that rapidly changes in the face of a fast-evolving industry is exactly zero of those things, and its value will be similarly small and round. It's merely a reference non-implementation, and since we're talking about browsers, we already have several candidates for being an actual reference implementation, just as we did in the old IE vs. Netscape days, when we learned for the first time why formal standards matter. Alas, as George Santayana observed a century ago, those who cannot remember the past are condemned to repeat it.

yuhong 15y ago
Well, ESR releases of Firefox are one thing, but browsers do not actually implement different HTML versions separately or as one unit.
MatthewPhillips 15y ago
I can't think of any instances where this happened. The closest I can think of is Web SQL, but that was still in a working group when it was canned. Developers who work on in-progress browser features are aware of the risk, or at least should be.