On 10/29/2014 09:14 PM, Sam Ruby wrote:
On 10/29/14 5:55 PM, Valentin Gosu wrote:
On 30 October 2014 02:47, Sam Ruby <[email protected]
<mailto:[email protected]>> wrote:
My assertion is that the steps I have provided cover these exact
same corner cases and parsing peculiarities, and I have running code
and published test results to back up my assertion. Can you
identify a single corner case that isn't covered (other than the
encoding changes that I already have noted as todos in the document)?
One I can think of is
var url = new URL("http://localhost/";);
url.host = "example.com:1234#hello <http://example.com:1234#hello>"
//hostname = example.com <http://example.com>, port = 1234
or even
url.host = "example.com:abcd" // hostname = example.com
<http://example.com>, port = -1
The intertwingly implementation doesn't explicitly cover this case from
what I can tell (althought the end result might be the same).
Todo added:
https://github.com/rubys/url/commit/484bdec6a51ed7c7956f225ba5e8e28151ac1129
I've finally gotten around to implementing setters. Early results are
available here:
http://intertwingly.net/projects/pegurl/url.html#setter-rules
I'm surprised that some setters accept trailing garbage, some percent
encode it, and some truncate it. I hope that the proposed spec text
makes this clearer.
I have an initial set of test cases:
https://github.com/rubys/url/blob/peg.js/reference-implementation/test/urlsettest.yml
Feel free to suggest more.
I've also produced a demo of a reference implementation that shows the
rules defined by this spec in action, and compares it to what the
browser does:
http://intertwingly.net/projects/pegurl/liveview3.html
Future plans are to compare browser results against the spec and against
each other.
- Sam Ruby
