[whatwg] Registering protocol handlers

[Ian Hickson]

On Mon, 24 Apr 2006, Peter Hall wrote:
>
> I meant a typo in the content type, not the URI.

In most cases that would just end up being ignored, and in any case this 
would be caught early in testing. And again, users wouldn't understand the 
concept of unregister-register.


> Perhaps there is no really good use case. But it feels right that, if 
> you do something, there should be a way to undo it. I certainly can't 
> see a reason why the ability to unregister would be detrimental or 
> dangerous

Any new feature has costs -- in no particular order: cost of 
implementation, cost of testing, cost of shipping the feature, cost of 
documentation, cost of writing the specification, cost of testing the 
specification, cost to authors of learning the new features, cost of 
authors to wade through the new features when looking for the other 
features, cost to the user of having to deal with the UI of the new 
feature, and so forth.

In this particular case, there is also the concern of security (sites 
removing other sites' URIs), which would require careful testing, and the 
concern that the user wouldn't understand what was going on.

To add the feature we have to be sure it is a good idea that will be used, 
otherwise the cost is not worth it.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'