Comment 90 for bug 294712

(In reply to comment #85)
> "content policy" does nothing with hostnames. A particular policy
> implementation (like nsContentBlocker) might do something.

I honestly don't know how or where it is done, my testing was litmus-style. I tried blocking an image on google.co.uk and it worked, not sure how though. I don't see any Add or Remove method defined in nsContentBlocker so I'm not sure where that gets implemented. I do see an implementation in nsContentPolicy though...

I think my testing was flawed anyways, it seems that all that is stripped is the www at the beginning of the url. I'll add that to my implementation with the next patch, that should make us compatible. Sorry for the confusion.

> > doing the same thing TestPermission does of culling between the second-to-last
> > and last '.' won't work for foo.bar.co.uk (or any other international suffix).
>
> Yet this is what nsContentBlocker does, right?

Only for BEHAVIOR_NOFOREIGN so I can't copy that for my use-case. I think I have this figured out anyways, so I'll try to be as compliant as I can.

> > As to the scheme checking, I can do that, but I just use GetHost, which is
> > pretty much the same thing.
>
> Uh... how could it possibly be the same thing?

I don't know of any other scheme that GetHost works for, is there anything else?

> > Do you want me to do that here?
>
> I honestly don't care; I just want the code to be internally consistent. You
> might want to not change the existing performance-sensitive codepath as part of
> a bigger patch, though.

Problem is, I need the host anyways for the entry which is why this was the fastest way for me to bail, ShouldLoad doesn't need the host. The reason I think it's faster anyhow is because ShoudLoad calls GetScheme first, then manually compares the string to http/https/ftp...

Unless there's a case where GetHost doesn't fail on non-{http|https|ftp} I'd like to use that.