Re: URL work in HTML 5

Anne,

I spent a quite a bit more time reading over your URL spec[1] and I have a
much better understanding of what you are trying to achieve. I also think
that your work is complementary to RFC 3986 and to many implementers it may
in-effect replace the need to refer to 3986.  I can see how specifying a
standard algorithm for parsing a URI could be beneficial for a particular to
a set of user-agents that are attempting to behave consistently.

However, for example, when it comes to how to deal with exception conditions
like spaces in URLs, I don't believe that all user-agents want to take same
approach to error handling that web browsers do.  In my own particular
scenario, I happen to know that the content providers, that my user-agent
deals with, use libraries that force URIs to comply with the 3986 spec.  If
my user agent were to have to receive a URI that contained an unescaped
space, I know something has gone horribly wrong and the last thing I would
want to do is silently fix it.  Different scenario, different needs.

So, I do see value in creating a URL processing specification for the "Web
Platform", I just don't think it should attempt to replace or overlap with
RFC 3986.   It should continue where 3986 stopped and apply its additional
constraints.  I am not familiar enough with the current willful violations
to know whether there are scenarios that just could not be handled without
changes to 3986, but I believe that would be a very interesting discussion
to have.

Maybe I am being completely naïve here, but from reading a number of threads
on various mailing lists that are talking about your URL spec, the sticking
point seems to be around your assertion that "The plan is to obsolete the
RFCs ".  I believe that from your perspective many implementers will not
need to use anything other than your URL parsing specification and so in
effect obsoletes the RFC.  However,  I see no need for you to duplicate the
effort of 3986 just so that you can layer some additional constraints on top
of it.  And secondly, 3986 needs to remain in order to provide guidelines to
those who choose not to conform to the Web Platform URL Parsing
Specification (forgive me for renaming your spec :-)).


Darrel


[1] http://url.spec.whatwg.org/

Received on Tuesday, 16 October 2012 05:38:38 UTC