Re: [AC] other issues

On Mon, 7 Aug 2006, Arthur Barstow wrote:
> 
> Presumably syntax errors such as this could be found before deployment. 
> Thus to facilitate future proofing, it seems like ignoring the invalid 
> value would be best (for example in case the syntax and/or semantics of 
> a pseudo-attribute's value changes in a subsequent version of the spec).

Quick note -- we should make sure that whatever the error-handling / 
fallback logic is, it defaults to secure rather than defaulting to allow.

In particular, if AC has a construct that allows a PI to override a 
previous PI to narrow the allowed domains, for instance, then a failure in 
the syntax of the second PI should lock out _all_ domains. (I haven't 
checked to see if that is an example that makes sense, but hopefully you 
understand the point I'm making.)

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

Received on Monday, 7 August 2006 19:27:09 UTC