[whatwg] 5.7.6 The application cache selection algorithm

[Ian Hickson]

On Thu, 29 Jan 2009, Michael Nordman wrote:
> >> >
> >> > Note that the "document-is-markup" flag will likely go away, and be 
> >> > replaced by just checking to see if there is a manifest (as it was 
> >> > before).
> >
> > The reason this particular thing might change is that I checked it in 
> > yesterday and this morning one of the Apple guys complained to me on 
> > IRC about it, pointing out it was a change from the earlier text and 
> > that he thought the earlier text made more sense. :-)
> 
> The behavior of HTML resources listed in the manifest but not having a 
> manifest= attribute would be very different if this change is adopted. 
> Unless somebody has learned something concrete from experience with this 
> system to indicate that this change is called for... probably should be 
> backed out.
> 
> I can imagine cases where this change would be difficult to deal with. 
> For example a static HTML page could not be incorporated into more than 
> one appcache. If you have a system that provides an appcache per user 
> (say by virtue of dynamically generated top level pages containing 
> manifest attributes), your only option would be to dynamically generate 
> all HTML resources within that cache to contain the appropiate manifest 
> attribute.

I've changed it so that XML and HTML files that don't have an explicit 
manifest will get fetched from an appcache if there is one from the same 
origin. (If someone _adds_ a manifest, then the cache will get updated 
with that new URL, and if the user goes back to that page, then it'll fail 
to load from the apocache since it is now foreign.)

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