Re: [dom3core] getAttribute

On Dec 2, 2005, at 11:40 AM, Robin Berjon wrote:

> So we have several hundred million copies of broken software out  
> there. In most cases this would indicate that changing the spec is  
> easier. However there also are, fewer but still quite a few, non- 
> broken implementations that could cause trouble if this change were  
> applied. Thankfully, all those on the former side use Javascript  
> (and perhaps VBScript), while all those on the other side use Java  
> (and sometimes C#). It would therefore seem to me that doing this  
> on a bindings level would work, with bindings not described in the  
> spec can decide for themselves. It's ugly, but it makes something  
> uglier go away.

I think you are making assumptions of things not in evidence, that  
all Javascript implementations are broken in this way and there is a  
clear dividing line.  In fact it is obvious that any number of prior  
versions of Javascript implementations implemented correctly and  
there are likely more in existence that are correctly implemented.   
We have also encountered enough cross-over during the years,  
including from Sun, that just citing a few broken cases is a long  
ways from authoritatively demonstrating that it is always true.

In fact there have been any number of Java implementations of HTML  
DOM exposed into Javascript implementations I have been aware of over  
the years.

All you believe you care about is a far cry from all and the triage  
may not be as easy as some may like to think, even if the broken  
cases were few.

If you can demonstrate it is true for every case that anyone cares  
about, you would have a starting point.

Ray

Received on Saturday, 3 December 2005 02:18:33 UTC