W3C home > Mailing lists > Public > uri@w3.org > October 2012

Re: automated testing for IRI (and URI) Specs?

From: Chris Weber <chris@lookout.net>
Date: Sat, 20 Oct 2012 16:14:52 -0700
Message-ID: <5083306C.4040403@lookout.net>
To: Larry Masinter <masinter@adobe.com>
CC: "public-iri@w3.org" <public-iri@w3.org>, "uri@w3.org" <uri@w3.org>, "Robin Berjon (robin@w3.org)" <robin@w3.org>, Rebecca Hauck <rhauck@adobe.com>
On 10/20/2012 1:08 AM, Larry Masinter wrote:
> Anyone know of any other existing IRI/URI/URL/LEIRI tests?
> For tests that push DNS we might need some other test framework
> (including an DNS server which will accept requests  <anything>.site.com  where anything is an arbitrary string, 
> and send request with Host header) so that I can test the IDNA conversion bits.

I have a compiled collection of about 500+ tests from Webkit, Julian, as
well as other sources available at: https://github.com/cweb/iri-tests.

I'd be happy to convert these into a JSON or some other format if it was
helpful.  I definitely agree that a DNS server, as well as a Web server,
would be useful for testing.  In my own setup I used:

1. The test cases linked to above.
2. A DNS server which would respond to *.foo.bar where * was an
arbitrary subdomain string as you noted.
3. A Web server which could:
 a) Serve each test case individually;
 b) Capture incoming HTTP requests based on each test case, allowing me
to analyze both the Host header as well as the path component in the GET

The test cases were served as both an href as well as an img src with
the goals being to see how Web browsers parsed these URLs in two ways:

1. How each test case was parsed and represented in the DOM properties
of the page.
2. How each test case resulted in an HTTP request, and how the
components of that HTTP request (host header and GET path) compared to
the parsed DOM (hostname and path) properties of the same test case.

Unfortunately my setup was not very portable, but could be made so with
some effort.  I actually did a little more than noted above, because I
ran a packet sniffer to see what the corresponding DNS requests looked
like as well.  I posted some of the early results up here:
as well as at Github.  I meant to clean things up, use a consistent DNS
scheme, define metadata better, test refs, define expected conditions,
etc. but didn't go that far.

I did notice some substantial differences across Web browsers, and also
found some instances where the HTTP request did not match the DOM
properties for the same URL.  But maybe these are mostly edge cases.

Best regards,
Chris Weber
Received on Saturday, 20 October 2012 23:15:09 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:16 UTC