W3C home > Mailing lists > Public > public-test-infra@w3.org > October to December 2011

RE: re: Does it pass?

From: Kris Krueger <krisk@microsoft.com>
Date: Thu, 3 Nov 2011 04:09:03 +0000
To: "Linss, Peter" <peter.linss@hp.com>, "public-test-infra@w3.org" <public-test-infra@w3.org>
Message-ID: <0A605FCDA3A4DC45A98B0DDD1C5351A7034C84C9@TK5EX14MBXW601.wingroup.windeploy.ntdev.microsoft.com>
The page shown at the session is below, feel free to pass this on to Russell.



-----Original Message-----
From: Linss, Peter [mailto:peter.linss@hp.com] 
Sent: Wednesday, November 02, 2011 6:14 PM
To: public-test-infra@w3.org
Subject: Fwd: re: Does it pass?

(meant to send this reply to the list...)

Begin forwarded message:

From: "Linss, Peter" <peter.linss@hp.com<mailto:peter.linss@hp.com>>
Subject: Re: Does it pass?
Date: November 2, 2011 2:55:10 PM PDT
To: Russell Berkoff <r.berkoff@sisa.samsung.com<mailto:r.berkoff@sisa.samsung.com>>

On Nov 2, 2011, at 2:36 PM, Russell Berkoff wrote:


The example of testing the dir="ltr" attribute was discussed at the plenary break-out session.

Say a user-agent implements the attribute but renders the result with  different kerning or just monospaces the result when compared to the sample representation.

Does the user-agent pass?

How would a test-framework differentiate between rendering variances allowed or disallowed by the standard (desirable or not)? How would a testing-framework extract character sequencing (in this example) from rendering variances even if pixel-level screen capture was possible.

SDOs only have the discretion to  test-to-spec rather than what might be expected by an end-user. SDOs also need test-frameworks that will be able to verify potentially thousands of test-pages without needing manual validation of the rendered output.

In general the idea with reference tests is to produce a reference page that _will_ render identically to the source if the test passes, and differently only if the test fails. If the test isn't actually testing other aspects, like kerning, both the test and the reference should be written so that there will be no differences in kerning between the two.

One way to achieve that is to use a special font. A number of the CSS tests, for example, use the Ahem font and references are built accordingly.

Received on Thursday, 3 November 2011 21:37:21 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:34:07 UTC