HTML WG F2F meeting at TPAC 2010 in Lyon (room 3B)

04-05 Nov 2010

See also: IRC log


intro from Alexey

Alexey: I call you about the organization chart
... I want to project the IANA slide that I think was skipped yesterday

(setting up projector)

(IETF and IANA is projected)

Alexey: IANA manages registries, and there are multiple entities that affect what IANA does
... If IETF adopts a procedure or defines a policy, IANA is required to follow it
... IANA does give input on what the policy should be
... IANA follows what IETF says in RFCs
... the other entity that affects IANA is the IAB (Internet Architecture Board) - talks to IANA about policy decisions like licensing
... IESG approves RFCs and so defines the formats, IAB controls the policy experts
... If people are unhappy with IANA policies they should not blame IANA - except in the case where IANA is slow in updating something

AVK: can blame them about format, URL persistence

Alexey: there is a document, RFC5226 which defines standard procedures for registries
... IETF can make any format that it wants, but there is a typical format for registries
... registries can have different policies, templates, levels of restrictiveness
... most permissive level is first come first serve
... examples include vendor names
... on the other end of the spectrum, the strictest ones require a standards track RFC
... in the middle is a procedure called "specification required"
... requires a stable specification from an IETF-recognized standards organization

IANA, rel, MIME, charset

HS: Is there an official definition of what is a recognized standards organization? there are different opinions

Alexey: no, it's not defined; people don't want to fix the list
... general criteria are: long established, stable document

HS: why is stability a requirement? if the software moves faster than the registry, then the registry is out of date

Alexey: depends on the registry - many registries are for developers
... for example, as a developer you may want to find all the link relations

AVK: but as a developer, I find current IANA registries useless
... wikipedia is a better reference for URI schemes than IANA is
... vetting by experts makes registries incomplete and inaccurate

HS: you said not just software implementors or others
... for years, image/svg+xml wasn't in the registry
... when Apple shipped MPEG-4, the type wasn't in the registry
... I can't think of any constituency for whom the registry says all that they want to know, or even close

AVK: apart from pedants, maybe

Alexey: a couple of comments on this
... different registries have different policies
... at the time when the registry was established, there was IETF consensus that this was the desired policy
... as time goes on, it may be that reality shows that a particular policy was too strict (or too permissive)
... maybe part of the answer is to revise the policy

HS: in the days of classic MacOS when Carbon was still used a lot, and you needed four char type and creator codes, it seemd that the value for those codes was smaller than the space for MIME types
... so you'd think you'd have a greater need than for MIME types to limit who can get what, but Apple operated a registry on first-come first-serve basis and nothing bad came out

<anne> MJS: you mentioned that it is possible to change the policy

<anne> ... assuming that some of the folks here are interested in a much more permissive policy

<anne> ... what would be the process to get the IETF to change

<anne> Alexey: talk to the AD and talk to other people to initiate discussion

<anne> Alexey: I'm happy to help with the progress

Alexey: the other half of the answer
... there is a reason there are expert reviews for some of the registries, like MIME types
... people do make stupid mistakes in MIME types, so there is an opportunities to fix this

HS: one of the supposed mistakes is using the text/* subtree for a lot of stuff, and there I would claim the mistake on the IETF side

AVK: what proportion of MIME types are not in use when they are registered? it seems like most of them already are deployed by the time you go to register them, so it might be too late to fix

Alexey: in the ideal world, people should ask experts up front

<Julian> !

Alexey: one example is that you can't use UTF-16 of textual types

HS: that's bogus

AVK: still insisting the case now is misguided

JR: one thing that Anne mentioned - some registries have a provisional system
... but not MIME types

Alexey: vendor prefix ones are first-come first-server

JR: other question -regarding the media type registration RFC, Larry has started discussing revising it in the TAG
... for example, people sniff for types - we could make that more robust

HS: I want to complain more about CR/LF
... the history of CR/LF restriction and the fact that text/* defaults to US-ASCII in the absence of charsets...
... this is an artifact of a leaky abstraction from SMTP
... US-ASCII default is a theoretical most prudent default from the time when in email there wasn't an obvious default
... but neither of those considerations apply to HTTP
... HTTP can send text that has line breaks that are not CR/LF
... in fact for HTML, LF-only is preferred
... it makes no sense to say that all these types like HTML, JavaScript and CSS are "wrong"
... instead it would make more sense to say that CR/LF does not apply to HTTP
... for some types, for historical reasons we need to default to Windows -1252 or UTF-8
... pretending these need to be registered under the application/* subtree doesn't help anyone
... it only serves the RFC canon that HTTP and SMTP match, but that doesn't help authors or implementors
... line breaks should be based on transport protocol
... types themselves should be able to define their default charset

JR: if you look at the thing that Larry brought to the TAG about MIME on the Web...
... he mentions all these problems
... line break thing doesn't make sense on the Web
... HTTP appears to use MIME, but doesn't, and doesn't need to
... charset is also an issue for HTTP
... conflict between MIME, HTTP and XML types on text/*

HS: I actually implement RFC2023
... I have a checkbox for saying ignore it

<anne> (There's a t-shirt saying "I support RFC 3023")

HS: if I shipped the validator without the "ignore it" box, people couldn't use the validator

JR: what's the default?

HS: defaults to supporting it

Alexey: comment on Web vs email - this needs to be discussed in IETF
... if Web requires modified version of MIME, let's do it
... there is a new WG in applications area

<anne> APPSAWG

<weinig> http://datatracker.ietf.org/wg/appsawg/charter/

HS: it feels frustrating to actually have to discuss this
... that people don't believe what they see on the web

AVK: the feeling is that the IETF is so much behind, and then we have to get in and tell the old timers what the new world looks like
... we're not sure it is worth our time
... we have moved on

Alexey: it is occasionally helpful to talk to people who designed the original
... especially when it comes to character set - I think there is agreement from the original author

AVK: I talked about some of the discussion about moving away from text/plain drafts, and people there express fear of Unicode....
... W3C is kind of slow too, but at least we think HTML and Unicode are ok

HS: well, W3C isn't ready to publish HTML5 as HTML5 yet

JR: IETF thinks HTML and Unicode are fine, just not for their documents

Alexey: there is provisional registration

AVK: for header fields, you need spec even for provisional
... person guarding the header field registry was too conservative

JR: does header name registry have a public mailing list
... registry lists should be public

Alexey: can you draw cases like this to my attention? it might be implementation of process failures

AVK: but if we look at URI schemes..

Alexey: it's hard for me to defend the people who designed the procedure
... there was a discussion about relaxing registration of certain types of URIs
... so we could register things like skype or yahoo IM

AVK: we are trying to register about: - there should be some registration pointing to the draft
... and for many headers, browsers have to know about them even if they are unregistered
... difficulty of using registry causes incentive to use X- names and just not registry

JR: one thing we should look at is accountability - there needs to be a public mailing list for header registration
... also Larry will join us to talk about IRI

AVK: I would rather just get rid of IANA and have a W3C registry, with a community-managed wiki

HS: to consider how the XHTML2 WG was doing things - at some point it was obvious that just giving feedback wasn't going to change the way they did things
... so instead of trying to change the way they did things, another group did something else, and that became the group people paid more attention to
... there is a feeling that fixing IANA is so difficult that it would just be easier to set up a wiki

AVK: we could just compete

Alexey: this is not helpful

AVK: I would like a registry that would tell me X-Frame-Options exists
... I don't think this will ever fly at IANA

HS: I have no experience of registration, but the language tag registry is a very positive role model

Alexey: when I talk to IANA, they listen

AVK: I think the problem is the process

Alexey: I can help you initiate changing the process

AVK: not sure I am interested in helping to fix the process if there is an easier path

HS: we should mention willful violations of the charset registry
... it would be useful for the main charset registry to be the place to go to find out what you need to implement
... the thing is that ISO-Latin1 should actually be interpreted as Windows-1252
... another example is that instead of Shift-JS you need to use the Microsoft tables not the ISO tables

LM: I note that my draft covers many of these issues

HS: not in this much detail; I will give feedback

<Julian> http://tools.ietf.org/html/draft-masinter-mime-web-info-01

LM: I hope in the cases where there are willful violations, that the right thing to do is to fix the registry

AVK: in the case of the charset registry, there might be a need for separate registries for Web clients vs other clients

HS: for example the Java platform uses the IANA names for charsets with their real meaning
... it would not be good to change Java, so the registry should include both sets of info
... JAva could add an API for Web content decoders

LM: I think this is a three-phase process
... (1) identify the problem
... (2) identify which things need to change (w/o being explicit about how)
... (3) then there needs to be action on the change
... I would like to identify the problem and the kinds of changes first
... only then decide whether to make a wiki, change the process, etc

AVK: if you are already working on this, then that's great

LM: I would be happy to have co-authors

Alexey: at minimum we should talk

LM: I think we should bring it into a working group or take it up as an action item
... MIME is a part of the Web architecture that we have adopted without adopting it

JR: we talked earlier about text/html and encoding

LM: again I think we should describe the problem first
... same thing might be said for URI schemes

HS: given last call schedule (1H2010), how realistic is it that changes of these magnitude could go through the IETF
... seems unlikely

LM: my view is that a W3C document entering LC can make reference to documents at similar or behind level of maturity
... they don't need to be final until you go to REC

MS: (explains W3C process)


HS: one reason I'm skeptical about the rate of change at IETF is the URL thing
... we had rules in the HTML5 spec abut transforming href values to IRIs
... it was argued that IRIbis was supposed to solve it
... I remember there was a schedule

LM: it's quite off

HS: at the date when there was supposed to be a deliverable, they haven't even started
... we shouldn't send things to the IETF to die
... I was really annoyed when I wanted to fix a bug relating to URL handling in Firefox and the spec did not have what was needed
... I think that for URLs the process has had it chance and din't deliver

RI: the original schedule was very aggressive and we never really expected meeting it

LM: it was wildly optimistic
... the problem with most standards activities is that there's nobody home except for people who showed up
... if you look at the archives, there was really a fallow period, but since then it is picking up
... meeting next week in beijing
... people who care about URLs in HTML should show up online

HS: there is also the problem that if people are already showing up in some venue, then moving the work to a different venue and then complaining that people didn't show up in the other venue is not productive

LM: the problem really is that what was in the HTML document before was wrong
... unfortunately there is complexity due to need to coordinate with IDNA and bidirectional IRIs

HS: you need something that takes a base IRI, a relative reference as UTF-16, and a charset, and you get a URI/IRI back
... my point is that the HTML spec doesn't need to deal with rendering any kind of address
... it just cares about resolution / parsing
... nothing about how to render an IRI
... what is required is someone writing down the real-world algorithm for this resolution thing
... and it needs to be somewhere that you can reference it

RI: if it were in the IRI specification would it be ok for you

HS: what I am annoyed about is that we had something that was right or fixable, was removed or delegated, and now we have to rewrite it
... I am now betting on Adam delivering it

JR: I would like to say one thing
... we need to find the right separation between things that are just part of the attribute and things that are part of the the resolving algorithm
... I think whitespace discarding is not part of the resolutions
... there might be a step before resolving that is part of extracting from an attribute

AVK: in the running code, whitespace stripping happens at the resolving end

LM: it would be nice if you could copy from the location bar into other apps

HS: we are not talking about the location bar

JR: what about space-separated lists of URLs

AVK: this is a different case

LM: motivation for trying to start the work in the IETF was to make sure that URLs in HTML and in other apps weren't different
... it is true that the work has been delayed, but activity has been restarted

Alexey: you need to open bugs

LM: Adam was at the last meeting
... there is an IETF document of how to do IETF document

HS: it's great the kinds of URLs that the web uses were the same as what other things use it, that would be great
... but the Web is constrained

JR: this was very useful, which I'm not sure was expected; we have another point about link relations, which is on the agenda


MS: in the future, we shouldn't delete things until the replacement is ready

LM: chairs from IRI working group are prepared to add an additional charter item

AVK: Adam is a bit reluctant to go back to the IETF

<anne> (that was my impression)

RI: it seems like there are discussions coming up in beijing where we need to be talking between the HTML WG and IETF

LM: editors will be remote, so remote participation might be good
... how about file: URLs

HS: they are not really on the Web
... best thing to do for USB key is relative URLs

<r12a> whether it's beijing or not, i think we need to find a way to pursue this dialog with HTML5 folks and chairs/editors of the IRI spec

RI: is something gonna happen
... action items?

LM: don't be skeptical - if you believe it will work

<scribe> ACTION: Henri to give feedback to Larry on MIME etc draft [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action01]

<scribe> ACTION: Anne to give Alexey info about registry problems [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action02]

<MikeSmith> started lunch break?

MikeSmith, we're about it

<MikeSmith> k

er, about to

session adjuourned

<anne> fwiw, testing was half an hour delayed

<anne> not sure if anyone is actually in the other room yet

<anne> but since you just signed in...

<Julian> isn't testing at 5pm (50 mins from now?)

<anne> no

<anne> it's a double block

<Julian> oh

<anne> yes

<anne> we are setting up

<anne> dbaron, ^^

<hsivonen> dbaron, we are in Rhone 3b

<hendry> scribenick hendry

<oedipus> scribenick: hendry


me: to find the connection type, it's not slow or rather blocking is it?

it's a fast operation Andrei: yes, we fire online when the type changes

type just caches last seen connection type


[ scribe apologies for pasting in wrong buffer ]

maciej: how to particpate in tasks tf, testing framework

<plh> kk: and goals for LC

kk: the TF meet every two weeks
... there is a wiki with schedule, there is a server with hg
... philippe has mirrored that work at http://dvcs.w3.org

<plh> --> http://dvcs.w3.org/hg/html/ HTML test suite repository

kk: same content on both servers

<plh> --> http://test.w3.org/html/ HTML Testing Area

kk: asking what to test ... localstorage, x-domain messaging, doing spec analysis
... looking at features which are shipping
... submitted some canvas tests

<plh> --> http://test.w3.org/html/tests/submission/PhilipTaylor/ Canvas test suite

kk: getElementsByClassname tests from Opera
... distinction between approved and un-approved tests

<plh> --> s/Philipp Taylor/Philip Taylor/

kk: bugzilla to process the test

<plh> --> http://test.w3.org/html/tests/harness/harness.htm Test harness

jonias: what is the harness ?

anne: same as XHR

kk: tests run automatically
... video tests is hard to automate
... self-describing test
... some exceptions that you can't poke in the OM and you can't test it

hsivonen: can you do some REFerence tests ?

jonas: yes, there are some things

kk: there are some things you can't test with REF tests, for e.g. Audio

hsivonen: multi-testing question

plh: some tests are manual and some tests are automatic

kk: existing tests not using the testharness, it might not be worth re-writing them

plh: it's a bug, it shows the buttons, though its automatic

kk: waits for 5 seconds before going to next test

maciej: this UI is broken

kk: can we get all the requirements up front ?
... esp we need a plan with REF tests

maciej: propsed categories; script driven, ref test, manual test
... too awkward with 100k tests ... takes too long to run

plh: the test can indicate itself, if it's manual or automatic

anne: if the test loads the test harness, we know it's an automatic test ( no need to categorise )

hsivonen: just have 3 directories

dbaron: you can harness the harness

kk: we should do it in one file

hsivonen: the easier way is to use directories

jonas: i don't care

maciej: text file is harder to maintain than a directory, not big deal either way

<plh> scripts/

<plh> reftests/

anne: we want directories for *types* of tests

<plh> manuals/

dbaron: painful to use dirs as metadata, as you may need to move them around

kk: maybe we will come up with a new dir in some months time, prefers a text file as it wont change location

jonas: bigger problem to have a function call when the test finishes so we don't have to wait 5 seconds after each one loads

anne: there is logic in the harness to handle this & async tests

hsivonen: [ didn't quite understand your implicit mochi test comment ]

<dbaron> plh: need a way to copy all the additional files that tests depend on

<hsivonen> I find that I almost always have to use the explicit finish function for scripted tests, so it's not a win to finish tests implicitly on onload

jonas: we need to somehow markup dependencies

sweinig: in the common case there will be no deps

hsivonen: should we decide whether to allow data URLs ?

anne: common resources makes sense

hsivonen: you want to use data URLs for near 0 load times

[ why does jonas use data URLs? didn't get his argument ]

kk: ie9 supports dataURIs
... might be a problem that browsers do not support dataURIs

jonas: we need to list our deps and assumptions
... can we assume browsers have ES5, foreach is nice

maceij: we should not use ES5 until it's widely implemented

jonas: queryselector test cases were held up by WebIDL

kk: e.g. of WebIDL false positive in canvas read only thing

jonas: do we have any existing docs of assumptions?

kk: there is just the source code
... can someone take an action to document them?

anne: read the XHR tests :-)

<krisk> testing wiki http://www.w3.org/html/wg/wiki/Testing

jonas: these tests are already in directories

kk: suggests documenting the tests in the wiki

hsivonen: ... something about re-writing the "mochi tests" ??

anne: i'm fine with re-writing / using another harness

kk: first anchor test is very simple, it's not hard to migrate to james's harness

jonas: make some requirements for making the tests portable between harnesses [ IIUC ]

hsivonen: something about integration layer, which allows reporting into your own system (thanks anne)

<plh> --> http://dvcs.w3.org/hg/html/ mercurial

plh: you can commit a test if you have a W3C account

dbaron: might need to be aware with hg's push caveats [ to plh ]

<plh> ACTION: plh to work with systeam to make sure we keep track of hg push [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action03]

maciej: not great security, since hg trusts the client's config WRT who wrote the patch

dbaron: you might want logs
... Mozilla have a tool called push-log for this problem

jonas: i can see now the tests are seperated by directory

<dbaron> The source for pushlog is in this hg repository: http://hg.mozilla.org/hgcustom/pushlog/

jonas: is there a description file ?

<anne> http://test.w3.org/html/tests/

<anne> http://test.w3.org/html/tests/harness/approvedtests.txt

kk: see http://test.w3.org/html/tests/harness/approvedtests.txt
... we will add extra info

jonas: remove domain so it's not server specific
... we have a test file per dir
... i want to walk this from the cmdline
... i want relative paths

kk: we might need some absolute stuff

jonas: i'm pulling via hg

kk: there is no absolute need for absolute urls

hsivonen: mochi-tests point to localhost

jonas: something clearly identifiable for a search & replace to get the tests working
... you can get different types of relative paths
... it's important that we can accomodate them in a "search & replace"
... we need to scale
... it's not workable to ban absolute paths

hsivonen: we need to document the "clearly identifiable" bit, like test.w3.org and test2.w3.org

jonas: we have to say it's OK to use abs paths

hsivonen: worried about some dir namespace collision
... get rid of prefixes

jonas: OK

<krisk> That is fine

kk: how to delimit the file ?

jonas: i don't care
... though, since it's hand-written, make it easy & little to type

sam: is there a preferred lengthmicroformats.org with CSS tests there was a wide range
... bad = long test & lots of permutations

hsivonen: we know a bad test when we see it

maceij: there is a fuzzy boundary

jonas: io bound if we have a million tests ... we need to keep it somewhat reasonable

sam: there are examples of tests that can be merged

adrian: there is a review process

kk: you could file a bug, raise issues

adrian: of course if it's approved, it doesn't mean it can't change again

sam: if all the tests pass, then the bugs are in the specs

kk: tests do content negotiation (canPlayTypepermanence) WRT choosing a codec the runtime supportS

hsivonen: mochi tests that we (mozilla) use, requires server side javascript

plh: was a lot of trouble already to support PHP for security reasons

sam: we have tests that use python, php, curl for certain load tests

<dom> (we evoked this in WebApps the other day; we can probably consider more server-side stuff at some point, but we need to need to have requirements documented earlier rather than ater)

<dom> (and please consider limiting the number of needed languages/platforms as much as possible)

jonas: we can generalise "slow load tests" so it doesn't neccessarily require PHP
... some security concerns here

plh: we need to review PHP files before they become live

jonas: we need it one the same server for same origin type cases

<dom> if same server == test.w3.org, that's part of the plan

hsivonen: we need a mechanism to load things slowly for example

<dom> (use a DTD for that)

hsivonen: avoid echo, we should return existing (approved) files

jonas: is there sensitive data WRT XSS-ing

plh: should be fine

<anne> safest might be w3test.org or some such

kk: what happens if 10 million tests are in the Q to be approved

dbaron: biggest risk is a test that claims to test something, but doesn't actually test it

sam: we should only accept tests that use the new harness
... the tests here are about testing regressions

kk: worried about approval rate, esp. if only he does it

plh: if a subset of tests are passed by everyone, they are probably good

anne: 1) is it good enough hsivonen 2) ... [ didn't get that ]

maceij: lets do a cost benefit analysis

<adam> Accidentally testing something that is not a requirement at all

maceij: 1st category testing undefined behaviour
... 2nd -- testing something contrary to a requirement
... -- at least one browser will fail this

[ can someone write what maceij said pls ? ]

scribe: 3rd cat testing something where it doesn't actually test it
... review should catch them all
... almost certain something will be wrong
... how much time should be spent on review versus benefit
... test approved == matches what the spec says

dbaron: from exp within CSS, review is more work than writing to test... so its not worth doing for an existing contributor

s/writing to test/writing the test/

dbaron: figure out why the test is failing sooner than later
... imp report: 1) run all tests 2) bug in test suite or in browser (v. time consuming)
... figure out WHY tests are failing

hsivonen: we should flag tests that fail in all browsers
... we can't assume the spec is neccessarily 100% correct

<hsivonen> we should flag tests that fail in 3 engines

maceij: low skilled tests don't need to be approved, better if everyone is just running them [ IIUC ]

anne: we should distribute the testing

maceij: don't have ref test when you could have a script test
... distributed test is more likely to succeed

hsivonen: do we have any way to feed the test info to the WHATWG HTML5 section info box things

kk: could be an admin problem if links change

<krisk> see http://test.w3.org/html/tests/approved/getElementsByClassName/001.htm for an example of a script based test

<freedom> nobody in 3B yet? there will be an EPUB related meeting right?

<oedipus> according to the agenda, EPUB discussion in 3B starting 8:30 french time http://www.w3.org/html/wg/wiki/TPAC_2010_Agenda#Room_3B__.28IRC_.23html-wg2.29

<mgylling> Reads 09:00 to me

<mgylling> To anybody who is physically there: does 3B have call-in facilities?

<oedipus> guess the first half hour will be spent in common again then breakout to 3B

<freedom> seems not

<freedom> I am in 3B physically now

<mgylling> freedom, thanks.


<MichaelC> scribe: Julian

ms: markus to give overview

mgylling: (remotely)

<mgylling> www.idpf.org

mgylling: epub standard for ebooks, around for several years, expanding in popularity, large adoption
... idpf.org
... based on xhtml, subsets defined
... current ebpub 2.0
... uses XHTML1.1 mod
... is a fileset, ZIP container, different document types
... container called OCF

<freedom> http://www.idpf.org/specs.htm

mgylling: some of the formats in epub defined by w3c
... some of the metadata formats owned by epub itself
... is undergoing rev to 3.0
...charter: update & alignment with modern web standard


use HTML5 as grammar

is not allowed by current specs but already happening

need to formalize & stabilize

on HTML5 vs XHTML5: epub decided to use X*

based on requirement for existing reading systems to be upgradeable

MS: asks about design philosophies
... drive spec based on what current UAs already can do?

mg: docs used to be static
... <script> SHOULD/MUST be ignored
... but scripting is going to be added
... problems with legacy readers
... and non-browser-based impls
... it's clear that this will be needed in the future

MS: devices coming to market with have full browser engines

Julian: usability of spec for being referenced


mg: not a problem yet
... we're not forking
... defining profiles and extensions, follow the HTML5 style

Julian: how does ext work for you?

mg: XHTML5 is supposed to allowed namespace-based extensibility

ms: feedback on this is welcome
... epub I18N requirements -> CSS WG -> vertical text support
... does not seem to affect HTML though
... is there something the HTML WG need to do?

mg: books / ebooks slightly different domain
... missing semantics for books
... distinguish node references and nodes
... skippability

page breaks

have looked at role attributes for extensibility

mjs: extending role not recommended because owned by aria
... needs coordination with PFWG
... maybe dedicated elements

or attributes

what affects rendering should be in HTML

mg: book semantics, chicago manual of style

in transcript, replace "node" by "note"

MC: asks about roles

MG: uses custom attributes

<MichaelC> Role attribute extensibility: http://www.w3.org/TR/role-attribute/#extending-the-collection-of-roles

MG: fastest way for now (own NS)

MC: role module *does* allow extensibility

<MikeSmith> RRSAgent:, make minutes

MC: PF and HTML need to coordinate on r@ole


<Zakim> MichaelC, you wanted to discuss role extensions, future aria, etc.

MG: ownership of @role

mjs: HTML defines @role by refererence to ARIA spec

MC: aria defines on HTML to define @role

<MichaelC> s/aria defines/aria depends/

mg: request to clarify the HTML spec wrt role extensibility

<fantasai> RRSAgent: make minutes

mg: on metadata in epub
... NCX doesn't have metadata at all anymore

<MichaelC> ARIA on host language role attribute

mg: core metadata will continue to come from outside HTML/head

<mjs> -> role attribute in HTML5: http://dev.w3.org/html5/spec/Overview.html#annotations-for-assistive-technology-products-aria

mg: reading systems need to get the metadata from the package file

HS: on role attribute

<fantasai> hsivonen: ARIA spec defines aria- attributes, but does not define role attributes

<fantasai> hsivonen: requires that a host language define a role attribute with certain characteristics

<fantasai> hsivonen: HTML5 tries to do this

<fantasai> hsivonen says something about tricky wordsmithing

<fantasai> hsivonen: Way forward would be to figure out roles that current AT vendors need (?) and define tokens for them, and have ARIA promise not to conflict

<fantasai> hsivonen: The role module spec relies on CURIEs for extensibility

<fantasai> hsivonen: ... not good for EPUB

<fantasai> hsivonen: I don't expect web engines to support CURIEs, relies on namespace stuff ... lookup DOM L3

<fantasai> hsivonen: Best way forward is to ask PF to set aside the names that you expect to use

<fantasai> hsivonen: Doesn't make sense to pretend different groups dont' know about each other

<fantasai> hsivonen: We're communicating, so let's coordinate.

<MichaelC> ARIA taxonomy

<fantasai> ?: I'm ok with approach Henri is suggesting, but coordination with PF is important sooner rather than later

<fantasai> MichaelC: Everything would have to fit into our taxonomy

<fantasai> hsivonen: Implementations don't care about the taxonomy, that's only to help out with spec design

<fantasai> hsivonen: If PF promises that this set of names is not going to be used, and picks different names if it decides to expand in that area, then we don't have to worry about all this extensibility stuff

<mjs> ack q+

<fantasai> MichaelC: For author understanding, we want to pick tokens that match the most appropriate terminology

<Zakim> MichaelC, you wanted to say if you want to follow the approach Henri suggests, should coordinate with PFWG sooner than later and to say ARIA roles are part of a taxonomy

<fantasai> hsivonen: They're just tokens, it doesn't really matter

<fantasai> mjs: Instead of debating in the abstract, let's just send the list of suggested roles to PF asap

<hsivonen> DOM 3 namespace lookup doesn't work for CURIEs in text/html DOMs, so don't expect browsers to implement CURIEs

<fantasai> mjs: If they don't like the tokens proposed, then they can respond about that.

<fantasai> mjs: I don't think this meta-conversation is getting us anywhere

<Zakim> Julian, you wanted to let Mike speak

<fantasai> hsivonen: I'd like to add a note about why CURIEs are bad idea in this space

<fantasai> hsivonen: So, frex, how Gecko exposes roles to interface to JAWS, Gecko picks the first role it recognizes and exposes that as the MSAA role

<hsivonen> IAccessible2

<fantasai> hsivonen: And then exposes the entire value of the role attribute as the xml-roles property in the iAccessible2 interface

<fantasai> hsivonen: It follows that the namespace mapping context of the CURIE binding context is not exposed at all

<MichaelC> scribe: fantasai

hsivonen: If you wanted to do something with CURIE, you wouldn't do CURIE processing.
... You would wind up exposing to JAWS the prefix and local name

<freedom> IAccessible2, http://www.linuxfoundation.org/collaborate/workgroups/accessibility/iaccessible2

hsivonen: Therefore I advise against relying on the mapping context, because the existing ... doesn't expose the mapping to IAccessible2 and therefore to JAWS

markus: Does Gecko expose the roles regardless of whether it recognizes it?

hsivonen: Yes. All the data is passed through, in case JAWS wants to violate ARIA and look at things itself.
... Gecko doesn't police whether JAWS follows ARIA spec

MikeSmith: I just wanted to state where things stand.
... It's not inconceivalbe that the language features you need for EPUB could be considered as native elements and attriutes to be added to HTML5 itself. It's not too late for that.
... It's not too late to ask, anyway.
... I'm sure we're going to get LC comments asking for new elements and attributes.
... There will be a lot of people who haven't looked at the spec yet, or want opportunity to have their request considered.
... Proper way to change the spec is file a bug against the spec.
... Cutoff for pre-LC was Oct1. Everything after that date will be considered an LC comment.
... I don't think that you should self-censor, and just assume there's no chance of getting any new language feature requests for native elements and attriutes considered.
... That's not what we want
... I don't want to say you have nothing to lose, because there's cost in time to everyone
... But something for EPUB to consider, whether you want to make requests for new elements/attributes.

<hsivonen> Gecko exposes the value of the role attribute to JAWS but not any kind of CURIE prefix mapping context, which mean using CURIEs wouldn't really work with the URL and you'd end up hard-coding a known prefix and the resolution to an absolute URI would be fiction

MikeSmith: Not mutually exclusive: could also pursue extensible approach, too

<hsivonen> thus bad idea to use CURIEs

MikeSmith: It's a good idea, although some things we need are likely to be considered out-of-scope for HTML5

Markus says something about e.g. notes

fantasai asks if that wouldn't be <aside>

mjs: Just want to reinforce Mike's comment that we would definitely like to hear all the requests, even though we are late in the game and probably aren't going to add major new feature.
... But requests that are modest in scope and important for a particular use case will be considered
... We're not 100% frozen yet, but in a few months we will be. So better to get those requests in now rather than later.
... Any other comments?

fantasai: Wouldn't notes be an <aside>?

Markus: Notes would be a subclass of <aside>

Markus says something about an href role

mjs: Talking about footnotes and end notes?

Markus: Yes. Need to distinguish those for formatting

MikeSmith: Don't we have a bug open on having more roles for <a>?

mjs: If particular semantic of linking to footnote or endnote might be more appropriate as a rel value

hsivonen: Maybe have a CSS pseudo-class detecting the note type from what the <a> points to instead of requiring author to specify

Markus: Reponse from EPUB authors say that overall, it's really good. There are a number of additions from XHML1 that we love.
... We're already very close to having it work for books, only a few minor concerns.
... So not looking for any major surgery here.

fantasai: I think they should define a microformat for subclassing notes.

hsivonen: HÃ¥kon and Bert already defined a microformat for books, although I don't think they addressed notes.

Bert: yes. A lot of that has been added to HTML5, though: <article>, <section>, etc.

mjs: HTML5 just recommends a plain <a>, with no distinguishing markup

hsivonen: footnotes are a thorny issue in CSS. Prince supports something, but it's not optimal
... I was reading Dante's Inferno in HTML5. It doesn't make any sense to read it without footnotes.

mjs: Yeah, I read a Terry Pratchett book that was supposed to have footnotes, but they were all endnotes and it didn't work so well

<Bert> Boom! (BOOk Microformat)

hsivonen: I think we should figure out the CSS layout model first, then fit the markup to that.
... If we come up with markup first, and it doesn't fit the CSS layout model, making it work in layout could become very complicated, involving many pseudo-classes, etc.

meeting closed?

<Bert> (Contrary to what I remembered, BOOM *does* have footnotes, not just sidenotes: <span class=footnote>)

discussion of role attributes

mjs: You need centralized extensibility for accessibility, so the a11y technology understands the roles

hsivonen: If you're on Windows, what FF can do is more than with the AS api on Mac

<MikeSmith> http://code.google.com/p/epub-revision/w/list

hsivonen: So maybe it's a bad idea to design stuff with the assumption that you have IAccessibible2 on Windows
... Alternatively, could consider it a bug that AS doesn't have this feature

<hsivonen> s/AS/AX/

anne: The only case you'd notice it is JAWS was updated before voiceover

hsivonen: I'm guessing the upgrade rate of JAWS is a non-issue in practice

<MikeSmith> http://code.google.com/p/epub-revision/wiki/Annotations

Julian: You might not believe how backwards some people are in upgrading their browser

hsivonen: Big parts of ARIA have been designed with the assumption of an enterprise stuck with IE7 for years after ARIA has been deployed in JAWS

<MikeSmith> http://code.google.com/p/epub-revision/wiki/DesignPrinciples

hsivonen: Design decisions make assumptions about which part of the system will be upgraded first. Might not have been the best design decisions.

<MikeSmith> http://code.google.com/p/epub-revision/wiki/HTML5Subsetting

fantasai: So is EPUB subsetting HTML5?

MikeSmith: not sure

mjs: Engines are unlikely to enforce any subsetting

fantasai: True, but such content could be non-conformant for EPUB 3.
... Not all EPUB implementations are based on browser engines

?: Are there many that are not?

fantasai: I know of at least two
... and I haven't actually looked into the issue

<kennyluck> fantasai: When I was at Tokyo, I found a EPUB implementation that implements CSS but not based on browser

<kennyluck> .. I also found one EPUB implementation that's not based on browser at all

<kennyluck> ... yet it renders vertical text quite nicely

<kennyluck> ... (It does not support CSS)

fantasai: uses effectively a UA stylesheet only

hsivonen: Are the CSS implementatiosn any good/

fantasai: Don't know, haven't done any testing

discussion of converting HTML5 to EPUB

would need to split into multiple files for EPUB impl's tiny brains :)

<mgylling> Yes, splitting files is done a lot due to memory constraints in certain handhelds

<mgylling> A popular one has a 300k limit IIRC

<MikeSmith> 12 minutes to caffeine

<freedom> which means EPUB doesn't encourage authors to write long chapters?

<mgylling> hehe, yes, need to keep it short ;)

<mgylling> I expect these max file size recommendations to be gone soon, just another generation shift needed in devices

<freedom> mg: do it, my iPhone 4 has 512MB now

<mgylling> freedom, right. Note that this is not spec restrictions; these are conventions that has arisen in the ecosystem

<freedom> OK, bad implementation, not bad spec

link relations

<scribe> ScribeNick: fantasai

mjs: Subtopics include
... Idea of using microformats
... another is that we have a number of specific issues

<mjs> http://www.w3.org/html/wg/tracker/issues/124

<mjs> http://www.w3.org/html/wg/tracker/issues/127

<mjs> http://www.w3.org/html/wg/tracker/issues/118

<mjs> http://www.w3.org/html/wg/tracker/issues/119

mjs summarizes the open issues

mjs: Does anyone else have other subtopics?

<adam> *u must be dozing off*

<anne> no kidding

<Zakim> MikeSmith, you wanted to show XPointer registry and to discuss potential need for a role registry similar to need for a rel registry

MikeSmith: Somehow I ended up the one responsible for registering all link relations for HTML5
... So, I guess I can put some kind of report on that? What should I be doing.

Julian: Let's start with a description of .. right now
... I'll summarize where IETF is right now.
... It all started with realization that HTTP has a Link header that's supposed to be equivalent to Link element in HTML
... And that there are documents on the web which are not HTML and for which it would be useful to expose linking
... Lots of people think it would be a good way of expressing link semantics independently of HTML
... So Mark Nottingham started on the work of writing a new def of Link in HTTP
... And establishing a registry that could be used in HTML as well, but would not necessarily be used in HTML
... The IANA registry also includes the link relations registry that was established for the Atom feed format, which is similar but not identical to HTML.
... So there are overlapps, but it included syndication-related things and not everything that HTML has
... So there was lots of discussion on procedural things, and licensing of the registry.
... Can talk about that later.
... Took a long time for spec to come out, but has finally been published.

<Julian> http://greenbytes.de/tech/webdav/rfc5988.html

Julian: That's a very old style: you send an email to an IETF list, and a group of designated experts to register that or ask questions.

<Julian> http://paramsr.us/link-relation-types/

Julian: Mark has started making this more modern by, first of all, providing a web page explaining how to register, has a template to help with you write the registration and submit for you to the mailing list

<Julian> http://paramsr.us/tracker/

Julian: The designated experts now also has an issue tracker
... So people can watch where there registration requests are progressing
... Makes the IANA process a bit more pleasant

<Julian> http://www.iana.org/assignments/link-relations/link-relations.xhtml

Julian: Here's the registry riht now
... This contains link relations defined in Atom, Atom extensions, and HTML4
... and some parts for HTML5

<Julian> https://www.ietf.org/mailman/listinfo/link-relations

hsivonen: ? has been recognized as an entity that has reasonable ? measures in place
... It seems that the domain name is owned by ???
... as an individual
... And whatwg.org is also owned by an individual

Julian: I'm not sure how that affects our impression of whether microformats.org is stable or not

<MikeSmith> s/???/Rohit Khare/

mjs: My biggest disappointment about the RFC is that it doesn't have provisions for individual registrations
... It would be useful to have a central repository where all of these can be listed so people know what's in use, even if it doesn't have a formal spec
... I think Mark should make a provisional registry.
... Mark said the registry would be so lightweight it wouldn't be necessary
... But that has not proven to be true.

<hsivonen> moreover, even proven to be false

Julian: We have provisional registries in other IANA things, and nobody's used them.

<MikeSmith> http://www.ietf.org/mail-archive/web/link-relations/current/threads.html

mjs: I think if you find something that's almost never used, then creating something that has higher barrier to entry, then creating something with a higher barrier to entry isn't going to increase use

Julian: People don't use provisional registries because they don't care enough.

mjs: microformats.org list has even lower barrier to entry, and it is used

Julian: One difference between IANA registry and wiki page is that wiki is completely HTML focused
... So they don't consider relations among other formats other than HTML
... They don't think about use on PDF or video

mjs: Most people invent link relations for HTML. I don't think it makes sense to force them to address these abstract link uses that may or may not be practical.
... It makes more sense to me to provisionally register the link relations, and then encourage them to think about generalizing to other formats.

hsivonen: It might be not about people not caring, but about provisional registration being dysfunctional
... I also agree with mjs that in some cases people don't care about nonHTML use cases. In that case we should just do HTML.

Julian: we talked about ... provisional registry [that hsivonen mentioned] yesterday, and I totally agree this problem needs to be investigated.
... I think we try.
... I think we should try to encourage people to think of link relations applied to non-HTML content

mjs: I think encouragement is fine. But if encouragement fails, what happens? Should the link relation then be undocumented because encouragement was unsuccessful?

Julian: ... nobody's mailed a link relation and asked designated experts to help make the link relation more generic

mjs: You've raised the barrier by tring to make it generic, the person doesn't care about making it generic, so it ends up being unregistered

anne: You don't need that to get it in the registry, but to get it endorsed

hsivonen relates hixie's experience with trying to register a link relation

hsivonen: If what hixie wrote wasn't enough, then I think we have a problem.

Julian: My point of view was that he didn't seriously try. He wanted to prove it didn't work.
... I don't think it will be productive to continue on this path.

mjs: When I looked at the original templates hixie submitted and compared them to what the RFC said, I couldn't see any mechanical procedure that determined they failed to qualify
... So it seems anyone trying to register would require multiple email go-around
... Same problems result in failure to register MIME types and URL schemes

MikeSmith: I have been going through the process of making requests using the mandated procedures

<MikeSmith> http://www.ietf.org/mail-archive/web/link-relations/current/threads.html

MikeSmith: You can see there the discussions about the registry
... It does take multiple go-arounds in email for these.
... One is for some of the link relation names or types, they are already being used in other contexts
... One of those was 'search'.
... If you look at that, it was specified somewhere else.
... Regardless of how you do this, there has to be some discussion about what this description should say
... I don't see any way to get around that, if you have multiple ppl want to define the same thing.
... Other issues were with how it's defined in the spec itself.
... 'up' is one of those. Had to go back to WG and get a resolution for it
... .. Maciej... having to change the description of the link relation so that it's more generic, and less about HTML
... I'm not thrilled with that.
... Don't really care about doing that at this point in the procedure.

<hsivonen> (one of the top Google hits for the metaphor is from one of our co-chairs: http://intertwingly.net/blog/2005/05/11/Fetch-Me-A-Rock )

MikeSmith: I think many ppl are not going to be thrilled about changing what they think is a perfectly reasonable discription of their use case to handle some speculative use cases
... That's alwasy going to be a troublesome thing for someone to do


MikeSmith: In the spirit of going through the procedure and taking it to the end to see if it ends up being something it works or not
... But I do think we have to keep open the possibility that we decide that it doesn't work.
... I don't think it's a given that just because it's an RFC and the registry exists, that we've commited to this is how we do it.

<MikeSmith> http://www.w3.org/2005/04/xpointer-schemes/

MikeSmith: I think it's still a possibility that this isn't working the way we would like it to work, let's try something else.
... There is something else, plh asked me to point out.
... Is the xpointer registry.

<anne> +1 to W3C doing web registires

MikeSmith: This is another way of registering something that is similar

<anne> s/registires/registries/

MikeSmith: I think the biggest ... difference between things that have been successfully regsitered
... and those that are still being reviewed
... i.e. provisionally registered
... All you need to do to request a provisional registration, you just start by typing in a name of some kind

it gives you a form asking for a description, and optionally a spec URL

MikeSmith: This is a middle ground between a wiki page


<hsivonen> This looks good to me

MikeSmith: At least it's got a form-driven interface
... I think this is a good middle ground
... If the IANA registry provided a way of doing this, I think that would be something we could agree on

Julian: IANA registry has something very similar
... The only thing is that instead of being automatically registered, it gets sent to the email list
... If we made a provisional registration out of the sumission, that would be the same.

<Julian> http://paramsr.us/tracker/

<anne> The requirements for XPointer are first-come-first-serve

Julian: and then someone on the mailing list to the tracker page

<anne> This is not at all the case for the link registry

<anne> well, the one the IETF/IANA uses

hsivonen: How do you know the tracker issue is filed and where that is?

Julian: You don't

?: Why can't you do a web-based form?

Julian: Can't do that in IANA. IANA doesn't have web-based forms. Lives in last century.
... The form that posts to email is a compromise.

hsivonen: So why does HTMLWG/W3C want to deal with an organization that lives in the last century

<weinig> s/?:/Sam

hsivonen: Instead of using xpointer registry code?

Julian: It depends on whether you think the link relations should be synced with other formats or not

sicking: Why couldn't you let W3C do the syncing to IANA?

MikeSmith: Before ? pointed out xpointer, I didn't know we did registries

mjs: Sounds like building a registry along the lines of xpointer would be a great idea

<MikeSmith> s/?/PLH/

mjs: Any volunteers to do that?
... write it up as a Change Proposal?
... It's a little past deadline, but since we have new info on the W3C registry option, would be a good thing to do

MikeSmith: Guess I should talk to plh about this.

hsivonen volunteers

MikeSmith: plh asked me to point out the open issue about Role
... We talked about it this morning. Similar potential need to have a role registry
... plh isn't sure xpointer way is the right way to go, but wanted us to be aware that it exists

anne: I think we should do role more centralized, because it affects implementations directly.

hsivonen: In last meeting I asked EPUB to ask PF to set aside some tokens for them once getting commitments from AT vendors that they will support these roles

mjs: Other things in HTML5 might benefit from this
... e.g. <meta> names
... There was a third thing

Julian: canvas context?

mjs: Seems more like role, in that it has implementation implications and should therefore be centralized

hsivonen: Yes. for role, e.g. you need coordination among AT vendors and browsers etc.
... Not good to have a registry. Rare to make a new role.
... PF should be able to set that aside without a formal process.

anne: Other one is meta http-equiv, which has a different namespace than meta name
... And canvas context, you do sorta need a place that says which are the contexts and which are compatible with which.
... Currently all are incompatbile, so not an issue now, but might change.

hsivonen: New canvas context is even rare


?: Still need a list of them

??: No, could just be defined by the specs that define them

hsivonen: I don't see this as being a problem right now.

<kennyluck> s/??/mjs/

hsivonen: There are three canvas contexts in the world, and one is proprietary

anne: we're removing them, 'cuz features have been added to 2d
... Might want a variant of WebGL that is compatible with 2D
... But still it's very limited

mjs: There's probably only a single-digit number of these, and should all go through HTMLWG anyways

fantasai: For link relations, seems like the idea is to have a provisional xpointer registry
... What about if someone wants to port a provisionally registered link rel to IANA, for more general use?


hsivonen: Dont't think we want to hijack Atom registrations

Julian: If we decide not to go with IANA registry, need to decide whether we want to continue with registration of HTML5 link relations in IANA

mjs: I think registering HTML5 link rels in IANA is unrelated to progress of HTML5
... It's not a requirement for us. It just makes the IANA registry more complete.

mjs expresses that he doesn't care whether MikeSmith finishes the registration since it's not required for HTML5

MikeSmith: It's not a lot of work, think it makes sense to finish offf.

mjs: what about the ones where the designated experts require changes to the definitions

MikeSmith: filed issues on that

mjs: For us, the importance of a registry is as an extension point.

sicking: Seems to me that the best caretakers of the link registry so far has been the microformats people
... So I want whatever solution we choose here to work for them.

mjs: Idea of using page on microformats wiki was proposed, but nobody's written up a change proposal for that either.
... Anyone want to volunteer to write that up?

sicking: Ok, I'll do it.

mjs: So post to the mailing list and say how long it will take you?
... I think we should make an exception here, because we have new information that will help us make a better decision

Julian: Microformats.org is not a new idea

sicking: New information is our experience with IANA

Julian: Half have gone through. A number are held on bugs being fixed in HTML
... Then we have to review the updated spec.

mjs: If the spec isn't updated, what happense?

Julian: We'd probably accept the registration anyway.

mjs: So why is the registration being held up?

Julian: If the description is updated at HTMl5, then the IANA registration would have to be updated multiple times.

hsivonen: Why is updating IANA registry multiple times a problem?

Julian: I don't think it makes a big difference either way

fantasai: Then I suggest you ask the IANA registers to finish the registration for any link relations that will be registered with the current text, and then update the registry when the problems they've pointed out have been addressed with updated text.

<scribe> ACTION: Julian to Ask the IANA designated experts if this would be an acceptable model [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action04]

<Julian> http://www.w3.org/html/wg/tracker/issues/127


Julian: ... Means in theory the semantic of the link relation can change depending on whether it's on <link> or <a>

<MikeSmith> trackbot, associate this channel with #html-wg

<trackbot> Sorry... I don't know anything about this channel

<trackbot> If you want to associate this channel with an existing Tracker, please say 'trackbot, associate this channel with #channel' (where #channel is the name of default channel for the group)

<MikeSmith> issue-127

<MikeSmith> issue-127?

<trackbot> Sorry... I don't know anything about this channel

Julian: I think the link relation should be defined the same for both, and the usage affect details like scope
... I think the section should be revised to not imply that rel values on <link> and <a> could be substantially different
... The IANA registry has an extension point so that each registration can have multiple columns

<MikeSmith> issue-127?

<trackbot> Sorry... I don't know anything about this channel

<kennyluck> trackbot, associate this channel with #html-wg

<trackbot> Associating this channel with #html-wg...

Julian: That was requested by Ian

<MikeSmith> issue-127?

<trackbot> ISSUE-127 -- Simplify characterization of link types -- raised

<trackbot> http://www.w3.org/html/wg/tracker/issues/127

Julian: E.g. to have a column that says whether the linked resource is required to be loaded, or just informational relation

<MikeSmith> ACTION: Julian to Ask the IANA designated experts if this would be an acceptable model [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action05]

<trackbot> Created ACTION-196 - Ask the IANA designated experts if this would be an acceptable model [on Julian Reschke - due 2010-11-12].

mjs: It seems that in practice the spec does what's requested, so it's more an editorial issue

Julian: This distinction applies both to the spec and also to the registry
... I don't think having the distinction in the registry is a good idea.
... We don't seem to have any good cases for that.
... The observation is, we currently have a table in the spec that has columns for effect on <link> and effect on <a> and <area>
... In this table, both are exactly the same
... except for two values, which in one column it's listed they're not allowed
... And in these case there are bugs on whether that distinction is a good idea.


fantasai: Setting stylesheet on <a> doesn't make sense to me

mjs: 'stylesheet' and 'icon' would have no effect outside <a>, even if we add them

Julian: ...
... We'll have to make a decision on that no matter where we put the registry. Defining things such that it's possible for relations to have a different deifnition on different elements is a bad idea.

mjs: ok

<kennyluck> s/<a>/<link>/

<Julian> http://www.w3.org/html/wg/tracker/issues/119

Julian: This is about the 'up' relation.
... Someone thought it would be nice to change the definition to allow repetition of 'up'
... to e.g. have 'up up' mean grandparent

mjs: That wouldn't work very well given the DOM api for rel, which lists unique tokens

fwiw, I agree this seems like an ill-fitted idea...

<Julian> http://www.w3.org/html/wg/tracker/issues/118

<anne> HTML5 says something different from HTML4?

<Julian> this is about navigational link relations that changed in HTML5, potentially changing existing content

hsivonen: fwiw, I think we should get rid of the up up up thing.
... It won't be supported in UI very well anyway

Julian: The use case given was to build a navigation tree in the UA
... But I think there are better ways to address that use case

hsivonen: When a browser user experience team wants to implement something, and asks for syntax for it, then we should conside rit.
... but at this point it just seems a theoretical idea
... So I would propose to just drop it

Julian: I'd like to ask the chairs to bundle the timing for these issues so they don't get too spread out

mjs: Could put them all together
... have been staggering them so you don't have to write proposals all at once


meeting closed

RRSAgent: make minutes

RRSAgent: make logs public

Testing 2

<anne> scribe: anne

MJS: Lets make a testcase in this session and submit it
... in the later half of this session

JS: I am willing to coming up with a format for tests
... and write a harness

<mjs> ACTION: sicking to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action06]

<trackbot> Sorry, couldn't find user - sicking

<mjs> ACTION: Sicking to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action07]

<trackbot> Sorry, couldn't find user - Sicking

trackbot, this is HTML WG

<trackbot> Sorry, anne, I don't understand 'trackbot, this is HTML WG'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<dbaron> trackbot, status

<trackbot> This channel is not configured

KK: I can update the wiki

<MikeSmith> trackbot, associate this channel with #html-wg

<trackbot> Associating this channel with #html-wg...

<scribe> ACTION: kris to update the wiki [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action08]

<trackbot> Created ACTION-199 - Update the wiki [on Kris Krueger - due 2010-11-12].

<scribe> ACTION: Sicking to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action09]

<trackbot> Sorry, couldn't find user - Sicking

<scribe> ACTION: jonas to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action10]

<trackbot> Sorry, couldn't find user - jonas

<sicking> gaah, i don't exist

<sicking> i irc, therefor i exist

KK: What about XSS issues?

PLH: I agree we cannot solve the XSS issues
... My goal is that we do not set up services on these domains
... so there is no problem, effectively

AVK: as long as w3.org does not document.domain we are fine, otherwise it might be safer to use w3test.org

MJS: There might be a problem in the future; everything should be safe if we do not use a subdomain

JS: I have an idea for non-automatible tests, but we can discuss that later
... The way I would like us to do new things is write tests in the new format if it is compatible with our features

MJS: We have a requirement for landing new features and we could require them to be written in the HTML format

AvK: We have used this format successfully already
... e.g. for server-sent events and XMLHttpRequest

MJS: one thing we might need to do is identify features in the specification which are not new but still need tests
... there is an HTML4 test suite

AvK: I do not think we should start from that

[people agree]

HS: How does updating work?

JS: We will have to figure it out

HS: for html5lib WebKit first lands in WebKit, I land first in html5lib

[HS implements for Gecko]

SW: We are not opposed to change

Pushing policy

AvK: I think if the test contributor is known the tests should just get in

JS: I do not agree, I think we should have a staging area

KK: I think so too

MJS: I think it makes more sense that the testing in browsers happens later and that tests should get automatically in

[scribe misses out on discussing Mozilla specifics]

staging area

KK: Basically you have a set of tests, and wait for them to be approved

MJS: What do you want the approver to actually do?

KK: cursory review

AB: I think it might be worth having almost automatic approval process
... for tests that pass in multiple user agents

MJS: why does there need to be this approval step? it will happen in distributed form anyway

AB: to increase the level of quality

MJS: it does not seem to happen now

AvK: agreed

DB: I am not sure that a approval process is good for known contributors

MJS: It seems like a waste of time of people to require people to manually run the tests in every browser before it is approved
... there will also be cases that fail in all browsers

DB: it seems you want a staging area because you want a known good set of tests
... an alternative approach is to ship a release, rather than delay on trunk

HS: not having a lot of process helped html5lib to move forward faster

MJS: with a release you know it does not get worse

KK: the idea of approved is that is done

AvK: so far that has not worked I think

MJS: I think you will always get more tests and with releases you know the delta and can review whether that is ok as you already know the previous release was ok

[something about multiple vendors contributing tests being awesome]

MJS: problematic tests can be removed from the release

<hsivonen> fantasai: Microsoft testa a lot of value combinations. Mozilla tests tricky edge cases.

<fantasai> fantasai: Different vendors take different approaches to testing, and thereby cover different aspects of the features.

<fantasai> fantasai: By putting them together you get a more comprehensive test suite

JS: if the release process does not work we can revise it

KK: i like to lock things done

DB: if browsers import the tests they will report the problems more quickly

KK: in the current model the test can be pulled right away

[mercurial haz magic]

JS: If I find something wrong should I fix the test and mail the list

KK: currently mail the list
... and open a bug

MJS: I think people who report the bug should be allowed to fix the test

AvK: you want to optimize for the case that is most common, and most common the bug reporter will be correct I think

DB: you should notify the person who wrote the test

JS: I am fine with attaching patches to bugs


<plh> --> http://lists.w3.org/Archives/Public/public-html-testsuite/2010Feb/0014.html Mercurial server

<dbaron> hg clone http://dvcs.w3.org/hg/html/

http://tc.labs.opera.com/apis/EventSource/eventsource-close.htm is an example of a test following the non-written guidelines

<dbaron> default-push = https://[USERNAME]@dvcs.w3.org/hg/html/

<dbaron> is a line that you'd want to add to .hg/hgrc after:

<dbaron> [paths]

<dbaron> default = http://dvcs.w3.org/hg/html/


<hsivonen> let's make one of these: http://ted.mielczarek.org/code/mozilla/mochitest-maker/

<hsivonen> that is, we should have a tool like that for the W3C harness

<krisk> see http://test.w3.org/html/tests/

<hsivonen> I'm already annoyed by having to wrap stuff in test()

<hsivonen> so I can't do ok(false, "FAIL!"); in scripts that aren't supposed to run

<plh> ACTION: Kris to add reftest handling in the test harness [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action11]

<trackbot> Created ACTION-200 - Add reftest handling in the test harness [on Kris Krueger - due 2010-11-12].

<krisk> http://test.w3.org/html/tests/approved/getElementsByClassName/001.htm uses a relative path

<hsivonen> https://developer.mozilla.org/en/Mercurial_Queues

<hsivonen> you'll really want to use MQ

Media Queries ftw

<krisk> http://www.w3.org/html/wg/wiki/Testing

<weinig> sicking: http://www.w3.org/html/wg/wiki/Testing

<plh> a reftest: http://test.w3.org/html/tests/submission/W3C/bidi-markup-export/html5-html/

<dbaron> trackbot, associate this channel with #html-wg

<trackbot> Associating this channel with #html-wg...

Summary of Action Items

[NEW] ACTION: Anne to give Alexey info about registry problems [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action02]
[NEW] ACTION: Henri to give feedback to Larry on MIME etc draft [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action01]
[NEW] ACTION: jonas to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action10]
[NEW] ACTION: Julian to Ask the IANA designated experts if this would be an acceptable model [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action04]
[NEW] ACTION: Julian to Ask the IANA designated experts if this would be an acceptable model [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action05]
[NEW] ACTION: Kris to add reftest handling in the test harness [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action11]
[NEW] ACTION: kris to update the wiki [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action08]
[NEW] ACTION: plh to work with systeam to make sure we keep track of hg push [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action03]
[NEW] ACTION: sicking to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action06]
[NEW] ACTION: Sicking to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action07]
[NEW] ACTION: Sicking to design a file format for describing tests, and to write a harness that will run the automated tests [recorded in http://www.w3.org/2010/11/04-html-wg2-minutes.html#action09]