10802 2010-09-29 11:05:00 +0000 Limit the number of identical items on the list of active formatting elements by removing previous duplicates when adding new items 2010-10-15 22:56:28 +0000 1 1 1 Unclassified HTML WG pre-LC1 HTML5 spec (editor: Ian Hickson) unspecified All All RESOLVED FIXED P1 critical --- 1 hsivonen ian ian james jonas mike public-html-admin public-html-wg-issue-tracking w3c public-html-bugzilla oldest_to_newest 39924 0 hsivonen 2010-09-29 11:05:00 +0000 Please add the scheme described in http://lists.w3.org/Archives/Public/public-html/2010Sep/0163.html to the spec. I'll suggest specific values for the tunable constants when I've analyzed the data Philip kindly provided on this topic. (I'm filing this bug now in order to have it on file before the deadline.) 40117 1 ian 2010-09-30 09:26:25 +0000 See the comment in bug 10801. I'm skeptical about specifying a specific algorithm here. 40999 2 ian 2010-10-12 07:48:48 +0000 EDITOR'S RESPONSE: This is an Editor's Response to your comment. If you are satisfied with this response, please change the state of this bug to CLOSED. If you have additional information and would like the editor to reconsider, please reopen this bug. If you would like to escalate the issue to the full HTML Working Group, please add the TrackerRequest keyword to this bug, and suggest title and text for the tracker issue; or you may create a tracker issue yourself, if you are able to do so. For more details, see this document: http://dev.w3.org/html5/decision-policy/decision-policy.html Status: Did Not Understand Request Change Description: no spec change Rationale: please see bug 10801 comment 1, but s/stack/list/. 41221 3 hsivonen 2010-10-13 12:50:25 +0000 Philip ran an instrumented parser over 422814 pages that parsed successfully. Here's an analysis of that data: maxNonFontDuplicates (cutoff: 0.999000) 0.9422: <= 0 0.9868: <= 1 0.9928: <= 2 0.9953: <= 3 0.9965: <= 4 0.9971: <= 5 0.9975: <= 6 0.9980: <= 7 0.9983: <= 8 0.9986: <= 9 0.9987: <= 10 0.9989: <= 11 Max: 7687 maxFontDuplicates (cutoff: 0.999000) 0.9468: <= 0 0.9826: <= 1 0.9890: <= 2 0.9918: <= 3 0.9933: <= 4 0.9943: <= 5 0.9950: <= 6 0.9956: <= 7 0.9960: <= 8 0.9966: <= 9 0.9969: <= 10 0.9973: <= 11 0.9975: <= 12 0.9977: <= 13 0.9978: <= 14 0.9980: <= 15 0.9981: <= 16 0.9982: <= 17 0.9982: <= 18 0.9985: <= 19 0.9986: <= 20 0.9986: <= 21 0.9987: <= 22 0.9987: <= 23 0.9988: <= 24 0.9988: <= 25 0.9988: <= 26 0.9989: <= 27 0.9989: <= 28 0.9990: <= 29 Max: 6829 This means that when adding a non-<font> formatting element to the list of formatting elements, on 94% of pages there was no identical element (element name and all attribute names and values matching) on the list *after the latest marker if any* already. On 99% of pages, there were 2 or fewer duplicates already on the list (after the latest marker if any). The worst case seen was 7687 duplicates. In the case of <font> duplicates, on 99% of pages, there were 3 or fewer duplicates already on the list (after the latest marker if any). The worst case seen was 6829 duplicates. So the worst cases are really crazy, so it makes sense to pick some limits. Furthermore, very low limits take care of the vast majority of cases. I'd be inclined not to differentiate between <font> and non-<font>, and simply allowing a maximum of two identical elements already on the list when adding a third. Again, please see http://lists.w3.org/Archives/Public/public-html/2010Sep/0163.html for how to deal with removing duplicates. I think it would make sense to put the limit in the spec, because it would suck if an HTML5-compliance scoring site like http://html5test.com/ put 4 identical formatting start tags in a test case and called an implementation non-conforming. 41227 4 hsivonen 2010-10-13 13:48:32 +0000 I did some further testing. I implemented my suggestion from this bug and bug 10801. The I extracted a list of pages that exceeded the limits from Philip's data. Then I loaded 24 such pages in the build with the limits in place and in another browser. I saw no breakage in the build with the limits. My choice of 24 pages wasn't random. I tried to pick pages where I could guess from the URL that they were unlikely to be filth I don't want to see. 41250 5 ian 2010-10-13 18:34:13 +0000 Could you elaborate on what limit you would like to see specified? 41358 6 hsivonen 2010-10-14 11:59:44 +0000 (In reply to comment #5) > Could you elaborate on what limit you would like to see specified? I thought I covered this in comment 3. If before adding an element to the list of active formatting elements, there are already more than 2 duplicates (after the last marker if any) of the element about to be added to the list, remove the earliest one. Then proceed with adding the element that you were about to add to the list. (AFAICT, "more than 2" can only be "3".) Additionally, please edit the AAA: In step #1 of the AAA, if the first "If there is no such node" check is true, abort the AAA and process the token according to the rules for "any other end tag token". (I could be persuaded that "more than 2" above should be "more than 1" instead.) 41445 7 ian 2010-10-15 22:56:04 +0000 EDITOR'S RESPONSE: This is an Editor's Response to your comment. If you are satisfied with this response, please change the state of this bug to CLOSED. If you have additional information and would like the editor to reconsider, please reopen this bug. If you would like to escalate the issue to the full HTML Working Group, please add the TrackerRequest keyword to this bug, and suggest title and text for the tracker issue; or you may create a tracker issue yourself, if you are able to do so. For more details, see this document: http://dev.w3.org/html5/decision-policy/decision-policy.html Status: Accepted Change Description: see diff given below Rationale: Reluctantly and obtusely concurred with reporter's comments. 41446 8 contributor 2010-10-15 22:56:28 +0000 Checked in as WHATWG revision r5638. Check-in comment: Add in some hard-coded limits for dealing with unclosed formatting elements to limit the explosive growth of the list of formatting elements in commonly-seen cases. http://html5.org/tools/web-apps-tracker?from=5637&to=5638