W3C

HTML WG f2f

04 Nov 2011 -- part 2


W3C Document license

Jeff: discussion of the document license predates my arrival at W3C

<pimpbot> bugmail: [Bug 14696] New: This is no longer true: "The end date is encoded as one day after the last date of the event because in the iCalendar format, end dates are exclusive, not inclusive" — so remove ...value="2007-10-20">19... <11http://lists.w3.org/Archives/Public/public-html-bugzilla/2011Nov/0299.html> 4** [Bug 14696] This is no longer true: "The end date is encoded as one day after the last date of the event because in the iCalendar format,

Jeff: so I urge others here to jump in and correct me if needed
... So, several years ago, the HTML WG determined that the W3C document license did not work for them
... and the group came up with a list of 11 use cases that were not possible with the current W3C document license
... the chairs of the HTML WG then brought that list to the W3C Team
... and the W3C Team took that list of use cases to the W3C Advisory Committee
... the W3C AC agreed with the majority of the use cases, but not with the use cases related to so-called "forking"
... which put the Team sort of between a rock and a hard place
... it turned out that coming up a a suitable license was very difficult
... this task was handed over to the W3C PSIG
... which did an analysis and proposed a license that they believed covered 9 of the 11 use cases
... then later in 2010 we decided that we needed to try again
... and in the end we came up with 3 candidate licenses
... the three licenses try to address the use cases and at the same time address the AC's concerns about "forking"
... I have not doubt that the PSIG left no stone unturned
... the chairs of the HTML WG created a poll
... in which they asked the group to consider the 3 PSIG-proposed licenses, and also 2 other more-permissive licenses
... the results of the poll were that a majority of the HTML WG members responded that they could not live with any of the 3 PSIG-proposed licenses
... for the other more-permissive licenses, the majority of the HTML WG responded to say that they could live with them, though there was a significant minority that said they could not
... So, where we are at now is that we do not have any plans to change the W3C document license
... but in another decision, we did great Community Groups with a more permissive document license

Marcos: we have seen an alternative solution, which is that editors publish their editor's drafts under "public domain" outside of the W3C

Jeff: we are on record as supporting a permissive license
... but the Membership told us by an overwhelming majority (80%) is that when you are on the W3C Rec-track, they feel that needs to not be forkable

Marcos: We have shared documents, and if the W3C doesn't provide a more permissive license, we are still going to be [publishing versions of the same specs outside of the W3C under a more permissive license]

<KevinMarks> you can't prevent forking by fiat

anne: it seems wrong to me that a secret club behind a Member-only wall say No to us and tell us what we can and cannot doe
... there is not opportunity for discourse there

<KevinMarks> all you can prevent is the spec representing implementations that have forked it

Jeff: I hear what you are saying. There is no "opaque wall" if you are part of a Member org, though I understand that it's different if you are an Invited Experts

[Jeff asks how many people in the room are IEs and how many are from Member orgs]

Jeff: The chartering of Activities go through the AC as well

anne: The issue is that many of the members in the AC are not even members of this group

tantek: as a rep of one of those member companies, I can sympathize
... I've also been an invited expert


.

scribe: there is the entire AC that votes on these issues
... but if you look at the participants in this group
... I can firmly say that as a Member, Mozilla does support a forking license for the spec

Jeff: part of my philosophy of change is to recognize where there are huge barriers to change, and to find other areas where the barriers are not huge, and work there
... so one of the reasons we create the Community Groups was to address this issue

anne: It would be nice if we were given reasons for why the AC said No
... the AC is advisory
... and if the Team feels that W3C should have a permissive license, the AC can be overruled

DaveSinger: I think one rationale was that they want a single specification to reference
... I don't think we would ever have a case where even if there is a moderately hostile fork, the W3C does not pursue enforcing the W3C document license
... so the horse has already left the barn

Tim: one argument is, I want people to say that they are putting their time and work into the [common place at W3C] where we have gotten together to do the work
... and we are all committed to working together at W3C to work on our specs here, not planning to then take them off somewhere else

<KevinMarks> the possibility of forking is what provides the social pressure to actually agree.

Tim: the case with forking of code is different, whereas with standards, [the argument for forking] does not hold as well
... I think it's important to not fork but also important to have the right to fork
... so these are conflicting needs
... I agree with Dave Singer that the license [does not have effect on what actually happens in practice]
... so I'm not sure we need to continue spending a lot more energy on this

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

TIm: we should be focusing on doing our actual spec work

Sam: This is perhaps something we can address with HTML.Next
... so maybe we can focus on that constructively

Kai: Companies have one very strong requirement, which is stability
... when you cannot rely on the standards to actually be standards, that is a huge issue
... W3C through its process guarantees stability
... my recommendation to my own company is that we have to completely ignore anything that comes out of the WHATWG
... just breaking out of the process because you don't like it is [not good]

Jeff: I do not work for any of the companies that didn't support forking
... but I know some of the reasons that we given
... I heard some companies saying that they do not want forking because we want one Web
... they said they don't want to see the Web get fractured
... for example, if some companies said they wanted a Web that has DRM features that are not part of a W3C standard
... I also agree with what David said, that maybe we need to do more educate, to help companies understand permissive licensing
... and I understand that Tim also supports doing such education

tantek: we all want one Web
... I think there are maybe some disagreements about how we get there
... in a past role, I was at Microsoft, and there was a standard called DVB-HTML
... and what happened to that standard?

Jeff: It died.

<KevinMarks> lets not even mention WAP

tantek: right. It's doomed to fail. In the long run , the one that survives in the one that's supported by reputation

<markw> CE-HTML is very much alive in OIPF, HbbTV etc.

tantek: I want to praise the W3C for the Community Groups
... it is our intent to develop the Fullscreen API in a Community Group with a forking license
... and it is more likely that for [new work we do] we are going to choose to pursue those in a Community Group

Jeff: one small clarification, I was neither taking a position for or against forking licenses


.

scribe: I was sharing what I heard
... this is consensus-driven org
... we saw a consensus to provide a more permissive license for Community Groups

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

scribe: but we did not yet see a consensus to provide a more permissive license for Working Groups
... but that could change and we could revisit it

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

<krisk> scribe: krisk

HTML.NEXT

<plh> --> http://www.w3.org/wiki/HTML/next HTML next wiki

<pimpbot> Title: HTML/next - W3C Wiki (at www.w3.org)

mikesmith: This meeting is to talk about what features we could add to the next version of html

samruby: the wiki is a location to store and track items we could potentially do in HTML.NEXT
... this was last updated in june

tantek: when we re-charted can we move the wiki?

paulcotton: Can you give a rational why to move?

tantek: it'll lower the cost to collaborate with other w3c groups (less logins..etc..)

mikesmith: all you need is a w3c user/password for w3c wiki's
... though html requires a HTML WG account to update the html wiki

karl: the main wiki was initially created for SW community, "forked" by QA WG, then more general including documentation. Some people were not comfortable using the wiki

samruby: Is the content updated on the HTML.NEXT wiki?

<MikeSmith> a shorter list of possible HTML.Next features from my recent presentation: http://www.w3.org/2011/Talks/TPAC/HTML5/#(15)

<pimpbot> Title: HTML WG Update (at www.w3.org)

davidslinger: these are all good items - though we should also start to add issues that we have with the current HTML5 spec to this wiki

Mikesmith: we have a bugzilla to track html.next bugs

<tantek> per mikesmith's point about w3c's wiki being more open/accessible - I see that as an advantage.

slinger: can we add a link to this from the wiki?

samruby: I'm not concerned with where it's at rather that we have it all in one place

<MikeSmith> http://www.w3.org/Bugs/Public/describecomponents.cgi?product=HTML.next

<pimpbot> Title: Components for HTML.next (at www.w3.org)

<MikeSmith> http://www.w3.org/Bugs/Public/buglist.cgi?product=HTML.next&component=default&resolution=---

<pimpbot> Title: Bug List (at www.w3.org)

jamesgraham: to add some of these we'll need components

<pimpbot> Title: Components for HTML.next (at www.w3.org)

shelly: I have another feature - make sure that all aria roles are in the spec

samruby: can you either open a bug or on the wik - indifferent which type, but it should exist

anne: components are in webapps, but it should use the parser so it should be html

plh: web intents - wants a new element, html may want to look into this as well for HTML.NEXT

samruby: they are creating an new element?

<tantek> let's not rathole on intents please

jamesgraham: if they are breaking the parser that seems like a very bad idea

<plh> --> http://webintents.org/ Web Intents

mikesmith: they don't have spec document, rather it's just a document

<tantek> mikesmith: I don't think he needed the intent element, he could have used the meta element

mikesmith: though they are using a new element it could easily be a meta element

samruby: bugzilla and wiki seems very light...

mikesmith: see the link that mike posted about potential html.next features

http://www.w3.org/2011/Talks/TPAC/HTML5/#(15)

<pimpbot> Title: HTML WG Update (at www.w3.org)

mikesmith: input mode?
... mobile browsers seem to have this use case
... one item that hixie removed from html5, is the datagrid
... some media items...
... playback statistics
... api media additions

samruby: not only datagrid is on the wiki

kimberly: if we start to talk about forms and mobile
... autocorrect is bad for password

slinger: this is already be taken care of another WG
... I don't think it's good to think about next features, rather we should be think about how to get the project launched in the best possible way

adrian: I like to note about a feature I added that's on the wiki (login)

anne: we should not be thinking about html.next as a big step, html5 was done incrementally
... I suspect html.next will also do the same thing...
... components may change how we think about interaction with existing and new elements created by libs
... we may want to bring some common lib elements back
... I want to see incremental progress

jamesgraham: we should publish one a year with a small number of items that are completed rather than a 15 yeare cycle

adrian: I agree with that notion

plh: I also agree with james

samruby: any more items?

paulcotton: let me replay some of the items that came up with at barcamp dealing with getting specs out faster

<pimpbot> bugmail: [Bug 14697] New: Harmonize roles with ARIA <11http://lists.w3.org/Archives/Public/public-html-bugzilla/2011Nov/0300.html>

paulcotton: I'd like to hear tantek view on doing future html work in a community group

<tantek> +1

paulcotton: for example webvtt

<tantek> I agree that we should consider community groups for HTML.next.

paulcotton: the organization is not getting patent protection because the spec takes so long
... by doing mondular, community groups, are ways to speed up the process

mikesmith: we have another HTML community group on editing, so we do have a precedence for doing html.next work in a community group

tantek: I want to follow up with paul's comment
... Mozilla agrees that doing HTML.NEXT work in a community group is the right place

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

<pimpbot> bugmail: [Bug 12715] When used to include data blocks (as opposed to scripts), the data must be embedded inline <11http://lists.w3.org/Archives/Public/public-html-bugzilla/2011Nov/0301.html>

<MikeSmith> minutes for the first part of this morning are here:

<MikeSmith> http://www.w3.org/2011/11/04-html-wg-minutes-part1.html

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 -- part 1 (at www.w3.org)

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

<pimpbot> planet: bhueppe: Journalism in the Open: The 2011/12 Knight-Mozilla Fellows announced <11http://blog.mozilla.com/blog/2011/11/04/journalism-in-the-open-the-201112-knight-mozilla-fellows-announced/>

<paulc> test

<pimpbot> bugmail: [Bug 14697] Harmonize roles with ARIA <11http://lists.w3.org/Archives/Public/public-html-bugzilla/2011Nov/0302.html>

Testing TF work and plans

<paulc> Mark: Could we actually outsource the generation of tests for HTML5

<eliot> scribe: eliot

<rubys> Art: I like the idea of more directed sponsorship

<rubys> Paulc: are we ready to outsource

KK: up front costs to hire omeone would be high. want to minimize that.

<rubys> James: we haven't built detailed test plans

KK: test reviews have to go on. not sure where they'd come from
... Where would the talent come from? But it's still a possibility.
... Set up requirements and make clear what has to be done.

PC: I have experience in this. XML Query test suite >25k tests
... missed. US Gov't wrote thousands of the tests
... They had some tooling that they fed the spe into to generate tests
... but also had peoplpe who were experts.

Look for XQuery test suite W3C

scribe: We seem to be concentrating on the browser venders. Have we asked everyone who's going to benefint from HTML5?

<paulc> http://dev.w3.org/2006/xquery-test-suite/PublicPagesStagingArea/

<pimpbot> Title: XML Query Test Suite (at dev.w3.org)

James: We focus a lot on numbers--necessary--but quality is most important.
... if we generate million tests but they don't find bugs, then effect is minimal.
... need the ones who will write evil tests.

Ryoske: we had refactoring tests. turned out needed 10 to the 16th tests. No way.
... important to catch the bugs

PC: XPath and XQuery tests went for core features first. Subsequent tests assumed the core tests worked.
... tests got more complex on the outer edges.

Russel: So when do I get a gold star? How many tests do I have to pass?

PC: W3C had workshop to see if they should do compliance testing
... given nonlegal structure it's unlikely that the W3C would do it. W3C is not in the certification business. Implies huge otlays of money.
... Do lots of validation testing. Company would use the test and make a claim to their customers that they pass a certain number of tests.

Mark Vickers: If we get these tests done, there's potential income as a service.

scribe: Lots of possible tests. If someone else runs the test, the legal responsibility moves.
... could provide regular income w/o w3c being in the certification business.

Russel: Even if w3c doesn't want to be in the business, there's a pressing need in various ecosystem to have cert programs.
... If w3c is going to present hundreds of thousand of tests, who sets up the buckets of tests?

PLH: are you suggestin w3c sharee that burden?

Russel: yes

PLH: w3c is not in the positio to do that. it will take several years. we have no plan.

Russel: for home network we do, for various components we go through UL.

PC: the logos are expensive to create. w3c is not in that business

Russell: We are expressing an ecosystem need

PC: Different WGs in the W3c have gotten through rec with minimum required tests
... divide to subfeatures. Create scenarios, and test for those. Get same result from multiple implementations, then we're good.
... WGs have set criteria low and sometimes high. If you set for certification, you set for infinity.

<anne> shall we write tests?

PC: it's important to talk about what needs to be done to get out of CR.

plh: we rely on members to submit tests. w3c doesn't have staff for this
... so far we are way far from the 200k numbers we'd need for html5

Wilhelm: Many cert programs have been deliberately harmful to the success of the web.
... one group made a test suite and froze the test cases. Years later another client came with those test cases. They didn't agree with the way the web had evolved.
... not sure how to solve it, but cert is dangerous to the web.

Mark: Might be some benefit for w3c providing test suites.
... 2nd milestone in which tests are completed?
... we also just need to be practical about certification. It's going to happen and the w3c should be involved.
... ideally at the infinite level, but more practical approach that grows over time. becomes more complete.
... can do this and potentially generate income for w3c.

kk: great conversation. I think in terms of the customer.
... if you put a logo on something and FB doesn't work, do they care?
... It's going to take a lot more than HTML. It's bigger that HTML. Bigger than w3c.
... if you have something on a device and you want that to work for years, it's going to take a lot.
... Great place to start the discussion.
... PLH to pass a hat for testing resources

Guiseppe: Not clear, is it a lack of money to close test cases?

pc: Maybe we're not organizing in the right direction.

plh: I told jeff i need 40 people and he said sure, give me the money.

DSinger: If you're doing HTML5, you can differentiate by saying you do more.
... if everyone else contributes tests and you don't you benefit.
... we urgently need to talk about how people who want to get certified can contribute

DBarron: CSS2.1, test suite served 2 purposes: validate spec. Let implementors converge.
... You have to stop at some point, but the test suite cont. to evolved

pc: You can find this on the web, the static version used for CR and the current contents of the repository.

DBarron: using test suites as a yard stick. the density of coverage per test varies greatly.
... one test could be more precise than 10k other tests.
... Very concerned when a test is used for more than did you pass or not?

Guiseppe: The are organizations interested. Sd organize and discuss: how to work w/o freezing a test case. and to discuss collaboration.

MVickers: There is another group. Doing everything we can to use HTML5., Driving the organization towards using w3c tests and not another standard.
... will follow up and arrange for a meeting. first with Jeff and then a subgroup of the HTML5 group.

kk: Members of the wg would help.

pc: plh is on point as the w3c person to take this idea and do something about it.

<jgraham> http://hoppipolla.co.uk/talks/testing/testing.html

<pimpbot> Title: Testing (at hoppipolla.co.uk)

pc: we'd like to get to some point where we have more people in the room to generate tests

Testing HTML

Link above is to slides James is going through

James: Who has used Mercurial? Afew. Written a test for a working group? Same few.

<pimpbot> Title: HTML WG f2f -- 04 Nov 2011 (at www.w3.org)

To get test in so you can edit, you clne the repository.

James: once you have the files, you can create a test
... a tests go into a directory /tests/submissions/name_of_vendor/featurename
... we create a directory. Use your favorite editor.
... create the test and then commit the test locally. It's just in your local repository, until you push.

<anne> "added 363 changesets with 2836 changes to 2081 files" o_O been a while since I looked at this

James: to figure out what to test....quite important.

<anne> how do I figure out what to test?

<anne> that is, figure out what is already tested

James: not testing conformance criteria for authors. Looking for implementor requirements.

<krisk> http://www.w3.org/TR/html5/tabular-data.html#tabular-data

<pimpbot> Title: 4.9 Tabular data HTML5 (at www.w3.org)

James: DOM interface has a bunch of properties. we might test for them...
... caption property on <table> isaimed at authors. Never look at text in green boxes.
... RFC2119 keyword is a tip.

"....MUST return the first caption element that is a child of the table. If there aren't any, it must return NULL." This is something we can test.

scribe: Writing JavaScript test: Testharness.js (link at end of slides)

<wilhelm> Documentation for testharness.js: http://w3c-test.org/resources/testharness.js

scribe: create a function called test
... straightforward construction.
... see "Test for the Table API" in slides for code.
... test for two caption in forst case and no captions in the second.

s//forst/first

scribe: that is simple, can be more complex. What happens if I insert a caption using the DOM, etc.
...SteveFaulkner: Is it always OK to leave out child elements and things?

James: when you're testing things you're not testing validity, so yes.

<krisk> If you don't have Hg installed http://mercurial.selenic.com/wiki/Download#Binary_packages

<pimpbot> Title: Download - Mercurial (at mercurial.selenic.com)

Wilhelm: Gives an indication of how many tests we need. One sentence and I can think of about ten things that need to be done.

Link is to Mercurial repository. Need to get Mercurial.

James: to find out what tests are needed, look at what's submitted, or reach out to testing task force
... to run the test, run it in the browser. It says pass or fail. You can add something to the assert message to make it more useful, but in general, pass/fail.
... We have a way to link tests to a section of the spec

kk: it's a big spec. It's hard to know where it's from

Peter: Can link to any anchor

James: Theere's a guide to handle the tests

<pimpbot> Title: A Method for Writing Testable Conformance Requirements (at www.w3.org)

James: if you're testing rendering, there's a process on the CSS page

Are the guidelines?

James: Not really

plh: do we need some

James: if the test is right that's good enough

How to address interdependencies?

James: General guideline is to not make test depend on things. Test the thing your testing only.
... you have to make some assumptions, but we try to make as few as possible.

Ryoska: Do we expect tests to restore states?

<yosuke> s/Ryoska/Ryosuke/

James: In general encouraged.

DBaron: I would say that even more strongly.

kk: there are some situations where you can't avoid it. You can comment that at the top of the test, if you need a clean cache or something like that.

Guiseppe: aree they run in the same order?

James: No guarantees.

kk: they should pass in any order. No order is assumed.

James: maybe now is a good time to stop, pick a part of the spec and write a test.

kk: Memory leak. We're supposed to validate that the normative parts of the spec work. Not whether a feature causes a memory leak.

<krisk> http://mercurial.selenic.com/wiki/Download#Binary_packages