CSS WG telcon

21 Jan 2009

See also: IRC log


_jdaggett, arronei, glazou, plinss, emilyw, Elika, Hakon, ChrisL, Bert, SteveZ, David_Baron, sylvaing, alexmog
Dave, Melinda




<jdaggett> +1.415.738.aaaa is jdaggett

<glazou> aaah

<glazou> is there someone else on the call ? It's all silent here

<glazou> no, a mac kbd :)

<jdaggett> heh

<glazou> I switched to the bright side of the force

<glazou> ;)

<jdaggett> very good!

<glazou> as soon as they're not ancient

<scribe> Scribe: Chris

<scribe> ScribeNick: ChrisL

CSS 2.1 Test suite

<fantasai> Arron: We're almost done on our end, and we're starting to upload files to the review server.

arron: nearing completion of our testsuite and bugun uploading to review server. approx 7k files need to be reviewed
... ned to share the effort

<dbaron> dbaron has joineed #css

Daniel: How long did that take to write?

Arron: About 3 months
... expect 5-6 months to review

Daniel: Coverage?

Arron: Not chapter 13, assume Melinda and HP have that covered


<fantasai> Glazou: This test suite is the highest priority of the working group

Daniel: Release of CSS 2.1 is highest charter proiority. Each organisation needs to contribute or else this will never happen

Hakon: This is a public review surely

Elika: Difficult to coordinate public review
... how to track stus of individual tests,track tests and comments and status.
... previously used mailing list, ok for low volume but not for a ton of tests
... needs a comment tracking system for individual tests
... Peter and I designed such a syste but its not been built

David: There are different types of review needed. Will not get all types on all tests

Daniel: Agree

David: Useful to have public review anythingthey want, but need comment tracking and fixing in a timely manner so we dont get duplicate comments too much

Arron: No problem with timely update to each file

Daniel: maybe a bugzilla with a bug per test

Elika: Its a lot of overhead. Wiki has less overhead per test, still a lot of hassle. Often need to comment on multiple tests
... tests come in sets, often a comment applies to multiple tests

Daniel: technical side is less important, needs to be efficient enough that all in WG can contribute to the reviews

David: What VCS is being used?
... maybe allow people to post patches using a distributed VCS

Elika: Its using subversion at the moment

Hakon: Could we use media wiki?

Daniel: Subversion has command line plus extensions for MacOS, Windows etc
... eg TortoiseSVN etc. Anyone should be able to use it

Chris: Yes, SVG is a good idea

Hakon: Need to be able to add comments without changingthe files themselves

<sylvaing> technology is not as hard as figuring out how and when a testcase has been reviewed imo

Peter: So we designed a system to do that, database

<szilles> Chris: tests are written in XML with list of statuses and authoring metadata

<szilles> Chris: we used Tracker to deal with comments, as Elika says this was inefficient for multitest comments

<szilles> Chris: we used CVS for the test cases

Hakon: Could do like sorce files with comments at the top to describe changes

Bert: might break some tests
... Need simple review. Most tests likely correct

Daniel: No, because from selectors we saw that the initially plausible tests ended up with many small changes

Hakon: Get www-style to help and get review comments on those which are wrong

David: Hard is everyone passes a test but its still wrong

Chris: Yes that sort of test is hard to catch

David: Need to link specific issues to affected tests

Steve: One way is to run them against implementations that are known bad. if they stil pass there is an issue

Elika: A wiki has too much overhead, and performance issues with that many tests.
... so options are bugzilla, tracker, or build something (if done soon)

David: Not convinced the mail list would fail

Elika: Its failing already to track all tests.

David; "has been reviewed already" is not really a binary state

Bert: Need some criterion to move it to the public site

David: Judgement call based on how long tests are around, comfort factor

Bert: Peole only say about tests with problems, not tests they thought were fine

<sylvaing> oops. never mind :)

Daniel: In Selectors test suite this was not an issue. if over several months no-one reported a problem, it still goes in the test suite

Chris: Sometimes approved tests have bugs filed on them and are withdrawn or fixed. its like shipping software

Arron: prefer bugzilla over wiki or a mail list
... can raise a bug on a range of tests eg background-*
... thats how we do it internally, in a database

Daniel: So do you open a new bug for all reviewed tests?

Arron: No, only failed ones

Daniel: But then you do not know which tests have even been looked at

Elika: Could use bugzilla to track problems but a wiki page where people can put their names against specific tests as being ok

Daniel: Good compromise I think. This is very important, everyone shoudl commit some time to review them

David: Would like to propose patches and commit fixes, do i need permission? Or do it and log that it was done?

Arron: need to be updated somehow

<Bert> (Wiki... or just an HTML page under CVS, the same CVS as the tests themselves?)

<glazou> isn't it http://dev.w3.org/cvsweb/CSS/CSS2.1-test-suite/ ?

Peter: Tests currently in SVN

<fantasai> http://test.csswg.org/svn/

<glazou> ok

David: Multiple reports of errors and the tests not changing is discouraging for further effort by the public

<plinss> or http://svn.csswg.org/test/

Hakon: yes. Also for 7k files we need a map overview of the overall state

<plinss> and http://svn.csswg.org/viewvc/

Hakon: very useful to have a tool, eg in OOXML and OF they had this same problem

Elika: We have plans for a system but lack time and budget to build it

Peter: i get time here and there but not enough

Hakon: Don't know anyone at Opera who could build that system

Daniel: Dont want to be stuck where we have tests but can't move forward as we want new tools. Have already reviewed tests with no svn, no wiki and it still worked. We need to start. in the worst case a maillost ise nough, with good subject lines

s/isenough/is enough/

Hakon: So if someone says a thousand tests have a given error

Arron: Then I would fix them and commit in the next few weeks

David: Please encourage bug reporters to submit patches as well

Hakon: OK as long as the patches are tracked

Elika: SVN does that

Hakon: So anyone on www-style can make patches?

Daniel: Make, or apply, patches?

Hakon: i meant commit

Elika: So they would need svn acces, not an issue

Hakon: if we get a crank then we can revert their changes

Elika: So send ok patches and you get SVN access?

David: Somewhere in that area. establishing trust after some good submission on the mailing list

Arron: no issue withanyone on the WG having SVN access straight away.

Daniel: Complex bugs would be discussed on the list first of course

Hakon: What we are not recording is the "ok" remarks

Chris: Elika mentioned a wiki page overview

Daniel: Mailing list is better, just say what has been looked at

Elika: hard to get an overview that way

Daniel: Not going to block the test suite forever. if a trusted review has happened then we go ahead. its trust, not number of reviews

hakon: But we need to track who has reviewed which test

Chris: yes you need to always log the revision number of each test reviewed

Elika so suppose we use the mail list for nowe tosend in reviews and Arron fixes them

Danile: OK as long as there are positive reviews as well as bug reports

Elika: OK so needs test names in subject line

Peter: Our system uses directory structure , svn allows files to be moved so a test can be moved to indicate status

Chris: Thats a big advantage of SVN over CVS

Elika: Moving files does break URIs

Daniel: Lets get started with the mailing list to get started

Peter: Will work on system as time allows

Hakon: Existing documentation on how to review?

<fantasai> http://wiki.csswg.org/test

Peter: See wiki

<dbaron> CSS2.1-test-suite/incoming/microsoft $ find . -name "*.xht" | wc -l

<dbaron> 6455

<dbaron> Are these the 7000 tests we're talking about?

Daniel: Please tell the mailing list where the documentation is
... Tests are all xht?

<arronei> Yes those are the beginning of the 7000 tests there are still more I need to upload

Elika: Thats the filename that Tantek and Hixie decided on. OK as long as server serves them with correct media type

<plinss> Browsable tests: http://test.csswg.org/source/

Berts question on new release

<dbaron> arronei, are the ~700 tests from a year or two ago included in those?

Bert: Current CR is 18 months od, lots of errata. good to ipublish a new version with all errata incliuded, and have an empty erratum list again. Suggest publish another CR, no status change just an update

Chris: +1 to that, much easier to read and avoids duplicate bug reports

Arron: Are there diffs?

Bert: Yes, member only but we could publish it. Not guaranteed to be valid HTML due to the tool used
... no problem

zaki, who is here?

Daniel: Jon had an agenda item on JLTF


<glazou> jdagget

John: Coold we get an afternoon for this?

(general agreement, Weds pm)

John: Also please send email if there are things that are needed

<jdaggett> http://tinyurl.com/baddirections

Chris: Is there a meeting registration form?


26. Turn left at Kalakaua Ave

1.9 mi

27. Kayak across the Pacific Ocean

Entering Japan

3,879 mi

28. Continue straight

June f2f location and date

Chris: Two different dates were being discussed. Which is the favoured one?

Daniel: June 3-5 or 24-26

Steve: June 3-5 not good for me

<jdaggett> jdaggett: either is fine

Hakon: Can't do second date

Hotels easier for the earlier dates, due to summer breaks

Daniel; Travel may be cheaper for earlier date too

Elika: Earlier date easier for me

Daniel: OK so we will go with 3-5 June. Sorry, Steve

Resolved: meeting is 3-5 June at Sophia hosted by ERCIM


<Bert> (Summer holidays in France start July 2. Pentecost is Jun 1.)

<plinss> +47 was howcome

<glazou> ChrisL: :I started keeping track of attendees

<fantasai> the other one was then probably me

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2009/01/22 14:47:15 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.133  of Date: 2008/01/18 18:48:51  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/ned/need/
Succeeded: s/Aprox/approx/
FAILED: s/Melin/Melin/
Succeeded: s/melin/Melin/
Succeeded: s/withas/with a /
Succeeded: s/stilla/still a/
Succeeded: s/sing/using/
Succeeded: s/ewe/we/
Succeeded: s/od/Or/
Succeeded: s/stat/state/
FAILED: s/isenough/is enough/
Succeeded: s/SVG/SVN/
Succeeded: s/SVG/SVN/
Succeeded: s/biga/big a/
Succeeded: s/Hakn/Hakon/
Found Scribe: Chris
Found ScribeNick: ChrisL
Default Present: jdaggett, arronei, glazou, plinss, emilyw, +1.253.250.aabb, +47.21.64.aacc, ChrisL, Bert, SteveZ, David_Baron, sylvaing, alexmog

WARNING: Replacing previous Present list. (Old list: _jdaggett, arronei, glazou, plinss, emilyw, +1.253.250.aabb, +47.21.64.aacc, ChrisL, Bert, SteveZ, David_Baron, sylvaing, alexmog)
Use 'Present+ ... ' if you meant to add people without replacing the list,
such as: <dbooth> Present+ _jdaggett, arronei, glazou, plinss, emilyw, +1.253.250.aabb, Hakon, ChrisL, Bert, SteveZ, David_Baron, sylvaing, alexmog

WARNING: Replacing previous Present list. (Old list: _jdaggett, arronei, glazou, plinss, emilyw, +1.253.250.aabb, Hakon, ChrisL, Bert, SteveZ, David_Baron, sylvaing, alexmog)
Use 'Present+ ... ' if you meant to add people without replacing the list,
such as: <dbooth> Present+ _jdaggett, arronei, glazou, plinss, emilyw, Elika, Hakon, ChrisL, Bert, SteveZ, David_Baron, sylvaing, alexmog

Present: _jdaggett arronei glazou plinss emilyw Elika Hakon ChrisL Bert SteveZ David_Baron sylvaing alexmog
Got date from IRC log name: 21 Jan 2009
Guessing minutes URL: http://www.w3.org/2009/01/21-css-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]