Talk:SC3-2-3-navigational-links-across-pages

• We should discuss how to deal with multiple pages

Wilco Fiers (talk) 13:35, 2 July 2015 (UTC)

Discussion auto-wcag workshop

• We need to add an assumption (for all test cases / in internal links): The tool does have a way to determine if a web pages belongs to a set of web pages.
• Does the test apply to a single web page or a sample of web pages?
• Do we need a threshold, e.g. Compare the list L1 to L2 ... Ln. If L1 matches ≥50% L2 ... Ln, return passed.

Closed issues

• Selector: "role=navigation" should be role="navigation" - done
• Step 4: Index of variable N is j. I assume the list assembled in Step 2 is referenced. - done

Frank Berker (talk) 14:17, 16 June 2015 (UTC)

• Step 3: Not sure if it is a fail of SC3.2.3 because both element and its ancestor do not have an id properties? What about a cannot tell when there is no id and a failed when ids are available but they don't match. - done
• Comment on Annikas's comment: We decided to cover it on another test case

--Kamyar Rasta (talk), 15 June 2015 (UTC)

• Selector: "Sometimes the active menu item (current page) is a text and not a link. If only one such item is contained within a list this list (menu) should still be selected.". Should this text necessarily be the menu item to the current page? if yes, how we can recognize that automatically.
• Step 2: this needs an "if" statement.

Kamyar Rasta (talk), 15 June 2015 (UTC)

• Selector: First the selector says "all list elements (ol and ul) that exclusively contain links" but then a number of exceptions are mentioned. This is confusing. Suggest to add a more formal description.
• Step 3: Typo: "debt" should be "depth".

Annika Nietzio (talk) 14:47, 12 March 2015 (UTC)

• Step 2: 'internal links' may be linked to the corresponding page.
• Step 2: The selector should be applied, before the existence of internal links is checked (to be in alignment with Step 1)

Frank Berker (talk) 08:01, 12 March 2015 (UTC)

• Selector: The instruction that the lists are flattened belongs to step 1, not the selector. Maybe this test could be simplified by selected any page and describing the construction of the list in step 1. If the list is empty, the test returns InApplicable.

Annika Nietzio (talk) 14:38, 26 February 2015 (UTC)

• Step 3: "Link text" should probably be defined. Does it include alternative textx for instance?

Wilco Fiers (talk) 10:01, 5 February 2015 (UTC)

• Consider this scenario: Page A has a navigation menu on the top. Page B has a menu with the same entries on the left hand side. According to step 3 these two pages would pass the test case. Is this the intended behaviour?

Annika Nietzio (talk) 14:47, 12 March 2015 (UTC)

• Comment on Annikas's comment: We decided to cover it on another test case

--Kamyar Rasta (talk) 15:30, 12 March 2015 (UTC)

• Step 2 and Step 3: 'lists P2L1 ... PNLN'. This expression reduces the understandability. We have single lists for the pages, so we can call them simply 'lists for P2 ... PN'.
• Step 3: The list of P1 could be taken as reference and be compared sequentially to the lists from P2 to PN. If one of them fails, F66 applies. If none fails, the SC is passed.
• Step 3: 'same relative order as all the previously compared lists' implies that all lists are checked repeatedly per currently selected list.

Frank Berker (talk) 08:01, 12 March 2015 (UTC)

• Assumptions: Remove placeholder "List of all assumptions"
• Selector: Improvement suggestion: list elements, where all items contain exactly one link and nothing else, or where all but one list items contain links.
• Selector: "flatten" should be "flattened"

Annika Nietzio (talk) 14:38, 26 February 2015 (UTC)

• Selector: Change "are using on web pages" to "are used on web pages"
• Step 2: "Make list" should be more explicit. I think it would help to say that this should be done the same way as in step 1

Wilco Fiers (talk) 10:01, 5 February 2015 (UTC)

• Step 2: Add definition of P2 .. PN. Are these pages determined by the list constructed in Step 1? (The assumption says "This test assumes that comparing a page to the pages that are linked from it is sufficient to judge the applicability of the success criterion.")
• Step 2: change outcome to "CantTell".
• Step 3: Why is "link text" in quotes? Does it need a definition?
• Step 3: Change "If all lists have the same relative order" to "If the links on P1 have the same relative order as the links on P2 .. PN".

--Annika Nietzio (talk) 14:48, 29 January 2015 (UTC)

Comment on Annika's suggestions: All suggestions have applied --Kamyar Rasta (talk) 09:33, 2 February 2015 (UTC)

• Selector: lists that "exclusively contain links". What happens if the current page is a item on the list but not a link? Should we allow list where all but one items are links?
• Step 1.3: Not all pages will have the same number of links. Use different indexes.
• Step 1.4: Comparing partial lists is not enough and not efficient either. Suggest to use only one list per page.
• Step 2: The control flow is flawed. Step 2 should not be inside the loop of step 1. In my understanding "return passed" will leave the loop. That means that the other lists are never tested.
• Step 1+2: "Check if both lists have the same set of items. If links on both lists are in the same relative order." Both lists have the same items and the same relative order, they also have the same order. That is not the intention of the test.

--Annika Nietzio (talk) 15:01, 11 December 2014 (UTC)

Comment on Annika's suggestions: all comments are applied, apart from suggestion about selector as we decided to ignore it on autowcag meeting. --Kamyar Rasta (talk) 09:33, 2 February 2015 (UTC)

• HL: this should be possible fully automatic. Detect all the sets lists of links of all pages. Combine the ones with the same destinations, and compare in these sets of the order of the destinations are the same.
• 50%: why 50%, if one menu fails, it should fail. This is only the order, it is not the question if every menu is exactly the same.

--Hanno (talk) 11:18, 10 September 2014‎ (UTC)

• This test is based only on F66 but G61 must also be taken into account. Especially for the selector: "List components that are repeated on each Web page in a set of Web pages". That means you have to look for components that are repeated (in this case navigation sections). Only repeated components are compared.
• The comparison should consider only pages which contain the same repeated element.
• In the instruction "more than 50% of all pages" it is not clear whether these 50% all have the same order, or if there can be different orders in that set.
• I am unsure how this test is to be interpreted: Assume there are two different orders of navigation sections, each occurring in 50% of the pages. Should this fail? What about 40% vs. 60%? 20% vs. 80%???
• Comment on HL's suggestion: It is not sufficient to compare the destinations. A link consists of both the link text and the destination.

--Annika Nietzio (talk) 13:43, 18 September 2014 (UTC)

• It is not clear what happens if there are several lists (ul, ol, nav). Are all the items added to one list? If yes, in which order? (Note that the DOM order might be different from the order in the HTML source).
• Add clarification what has to be compared (the link text or the href attributes or both?).

--Annika Nietzio (talk) 13:56, 23 October 2014 (UTC)

• How many menu's: The test no pass as soon as 1 menu has the same order. In reality there are multiple menu's: footer, login, main menu, sidebar. All menu links should be in the same relative order.

• Selector: We need a more defined selector. When it is a menu you might have html and images or '-' or '>' seperating the menu-links.

Also we should define what is not a menu. If you have links at the end of a text for further reading, that shouldn't count as a menu.

--Hanno Lans (talk) 15:01, 4 December 2014 (UTC)

• Comment on HL's suggestion (How many menu's): As we discussed last week, a common situation is when pages have a header and footer both contains navigational links and the header is different from the footer the order. I don't think the SC3-2-3 fail this situation. However, what you are proposed would fail that because you are demanding both of them have the same order.

• Comment on HL's suggestion (Selector): For the time being, we have assumed to use group of links as the most viable path to detect navigation mechanism for this test.

--Kamyar Rasta (talk) 14:08, 11 December 2014 (UTC)