Tests 4.32 onwards

OK, I've been through my tests and have edited the document Kai sent 
this afternoon accordingly (see attached).

In addition to the changes in the document, I have made some notes on my 
thinking below.

Until tomorrow...

Phil.

4.32 Suitable - I believe I've cleaned that up a little, basically 
setting the bar for a FAIL very high in an attempt to ensure consistency 
between evaluators.

4.33 Tab Order
Minor edits with some examples added for good measure

4.34 Tables Layout
Tidied this up. My worry here is that we're testing a balance between 
what is possible using CSS and what is practical in real life. I tried 
very hard not to use tables for layout on the fosi homepage - but ended 
up using 2 all the same. It wouldn't even pass the Basic test...

4.35 Tables Support
The only way I can see to resolve this is to send a request for a page 
that contains tables using the UA string of one or more devices known 
not to support them. Perhaps the emulator could include this functionality?

4.36 Testing
Are we constrained always to give a Pass/Fail/Warn? If not then I 
suggest the way to test this is to list the actual devices that were 
used (probably by means of a white space separated list of their Device 
Description URIs). It's not a test as such, but a way to report that 
tests have been carried out, which is what we want people to do with 
this one. So the 'Test Procedure' would be:

* List the identifiers of the devices on which the content was tested 
and rendered as intended
* List the devices on which he content was tested on which teh content 
was not rendered as intended

4.37 URIs
I've removed the example from here - yes, those things are true, but 
they don't actually exemplify the tests which seem pretty clear to me as 
they are.

Also, I've deleted the open issue as I'm not convinced that this is 
fully machine testable. I have some experience of this in the ICRA label 
generator.

Users are asked to enter the homepage URL. When the form is submitted, 
the system strips off the www subdomain where I can determine that it is 
not important. A simple test is to do a GET on the URI with and without 
the www and compare the content. If you get a 4xx message from the 
non-www, OK, the www is necessary, but you might get a 3xx or even a 200 
and still the content won't match byte for byte as the links and 
references within the content might be delivered in a dynamic process. 
The text could easily include something as simple as a PHP ?echo 
SERVER-NAME parameter which would change depending whether the www were 
included or not.

As a result I ended up saying that if the content returned from the two 
URIs differs by <= 95% - and that's a pretty arbitrary figure. The 
manual test is easier in this case!

4.38 Use of color

Generally tidied up but no substantial changes.