[whatwg] Section 5.7.3.3 Parsing cache manifests

[Ian Hickson]

On Tue, 30 Dec 2008, Alexey Proskuryakov wrote:
> 
> 2. When parsing fallback maps, there is a requirement to drop entries 
> whose namespaces are already in the mapping. I don't think that it is 
> helpful - fallback namespaces are matched by prefix, so it doesn't 
> resolve ambiguity. I suggest dropping this requirement, possibly 
> specifying that if the map is ambiguous, the first match wins.

It removes the need for preserving an ordering. Right now, the spec only 
needs to resolve ambiguities with respect to namespace length.

The only blackbox-testable difference between what the spec does now and 
what you describe (keeping the duplicates and resolving ambiguities by 
keeping the manifest order and using the first one) is that the fallback 
files that will never otherwise be used (the fallbacks for the duplicate 
entries) will be fetched (and could fail a cache update) in your 
implementation. That seems like unncessary work since it's a file that 
will never be used by the user agent.

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