- From: fantasai <fantasai.lists@inkedblade.net>
- Date: Wed, 23 Sep 2009 15:31:36 -0700
- To: www-style@w3.org
Summary:
- RESOLVED: Keep image-fit and image-position property names, explain rationale to SVG.
Also discussed test suites; conclusions follow:
- RESOLVED: New (more realistic) deadline for test submission for CSS2.1 is 15 Jan 2010.
After this deadline submitted tests will be accepted for interop testing,
but not for the CR->PR test reports.
- RESOLVED: After 15 Jan 2010, each vendor must run all submitted tests in the test
suite and report back on any problems with the tests. The deadline for the
first iteration of this is 15 February 2010. The deadline for the second
iteration is 15 March 2010.
- We will encourage and track individual human reviews of tests, and these tests will
have a higher confidence rating. However, once a contributor has demonstrated a
consistent level of competence, that contributor's test submissions will be included
in the above process even without individual test review.
- Goal is to have a preliminary test coverage report by TPAC so that we can identify
and plug holes in our test coverage.
Status report and work remaining:
- Arron and fantasai have organized the test suite repository. Expect to check in all
submitted tests within the next week or two.
- Repository currently has ~8000 tests; final estimate ~10,000 tests. dbaron points
out that many of the tests from MSFT are too divided and should be merged.
- Test suite build and indexing scripts need some work to handle directory restructuring.
(An HP contractor worked on this last year. Current code, along with a junk test
collection for testing, is in <http://test.csswg.org/svn/build-test/>. This will
build a test coverage report once set up correctly)
- We'll need the new browser test harness to be set up for production use after Jan 15th.
Unknown what is required for this; HP leveraged MWI harness to build a prototype, but
the system is currently unowned.
(Design docs, prototype, and current (PHP) code linked from
<http://wiki.csswg.org/test/harness>.)
====== Full minutes below ======
Present:
César Acebal
Tab Atkins
David Baron
Bert Bos
Elika Etemad
Daniel Glazman
Brad Kemper
Peter Linss
Steve Zilles
<RRSAgent> logging to http://www.w3.org/2009/09/23-CSS-irc
ScribeNick: fantasai
Agenda
------
<Bert> http://www.w3.org/mid/A56057A4896BEC4BA13EE4C45CEB4FED2B7F7D19E6@GVW0538EXC.americas.hpqcorp.net
Peter: Anything to add to agenda?
TPAC
----
Peter: We have a wiki page for proposed topics, please add topics
<glazou> TPAC wiki page http://wiki.csswg.org/planning/tpac-2009
Peter: There's a question of what we need for facilities
* dbaron hopes the parking at the meeting hotel is free
Bert: Only need to know whether we need a phone or a flip chart
Steve: Is there a whiteboard? We need something to write on
glazou reads out equip list: just chairs for members, network, and
LCD projector
Steve: So we need a flip chart
glazou: Do we need a phone?
Steve: Ask by email?
Tab: I would like to call in, as I can't make it in person
ACTION Daniel to ask how many members will need to call in
Daniel: Was thinking of arranging a meeting of present and former CSSWG
members for dinner on Sunday
Bert, Steve in favor
Daniel: So Sunday 7pm before TPAC in lobby of hotel
<Zakim> +mauro
Bert might not be there quite yet
<dbaron> arrivals to SFO from Europe are unlikely to get in that
late unless you have to change on the east coast
Peter: Any other items for TPAC?
<silence>
<dbaron> of course, if you're going to SJC, you probably have to
change somewhere...
Steve: Does anyone have plans for dinner on Monday night?
Steve: We can do that on the fly at the meeting
<dbaron> hopefully we'll have enough cars to transport people
<Bert> I won't have a car. (Plan to stay at conference hotel.)
* plinss should have a car (if HP will let me come...)
* glazou plinss, Bert, I just made request for flip chart + polycom
CSS2.1 Test Suite Deadlines
---------------------------
Peter: dbaron sent out an email about this
<plinss> http://lists.w3.org/Archives/Member/w3c-css-wg/2009JulSep/0101.html
dbaron: Back in June we agreed the deadline for submission of css2.1
tests would be last week
dbaron: I did submit a whole bunch of tests
dbaron: But it seems the deadline should be at a point after we have a
list of what we have
dbaron: Certainly getting a list of all the tests we have blocks getting
reviews
dbaron: for which we have another deadline in 3 months
Peter: Do we have more tests to submit, or are we just waiting to plug
in the holes?
fantasai: We also have e.g. bz, who wants to submit tests for issues
we haven't resolved yet
dbaron: We also have contributors who would want to submit tests, but
want to know what tests we don't have yet so they can write those
Peter: Sounds like the primary issue here is getting a single list of
what tests we have and what tests we need.
<dbaron> yep
Peter: How do we do that?
Steve: Does it make sense to have part of our F2F be a review of test
coverage?
fantasai: no
fantasai: On Monday, Arron and I did some organization of the test suite
repository.
fantasai: Next step is to check in all the tests that have been submitted.
fantasai: Then we need to index them.
fantasai: We have a script that indexes tests, but it currently it has some
limitations and needs some work
fantasai: Also need to check that the submitted tests have the relevant
metadata.
fantasai: We may have this done by TPAC, maybe not. Hope to have it done
at least by December.
glazou: How many tests do we have now?
<glazou> find . -type f | wc
<dbaron> or really: find . type f | wc -l
fantasai: I'm getting 16,000 files.
<glazou> find . -type f | grep -v svn | wc
Tab: You're getting a lot of files in ".svn" directories
<fantasai> ~8000 tests
fantasai: 8000 tests
fantasai: And I haven't even put in a lot of the tests submitted over the
past 6 months.
fantasai: So probably close to 10,000.
glazou: We need to release CSS 2.1 as REC in the charter timeline, and
the number of test reviews we have on the radar is scary.
fantasai: We should at least have all the tests needing review in one
place by TPAC.
fantasai: I'll try to get a coverage report by TPAC.
fantasai: At that point, we'll at least be able to see where we don't
have much coverage.
fantasai: The one thing there is that Hixie's tests haven't been checked
in yet, and they don't have the metadata.
fantasai: Arron has reviewed Hixie's tests and made a list of what metadata
they should have.
fantasai: If Arron could work on that we'd have a much more realistic summary.
fantasai: Actually, they might have the metadata. I don't remember.
peterl: Sounds like by TPAC timeframe, we'll have a better handle on where
we are.
peterl: Do we want to shift the deadline for contributions to...?
fantasai: What does the deadline mean?
dbaron: It means either we won't accept them until after implementation
reports, or we're branching the test suite
dbaron: That's what I think it should mean anyway
fantasai: After we do any implementation reports, or after we do the
implementation reports that pass
dbaron: after CSS2.1 reaches PR
Peter: When do we want the deadline?
fantasai suggests February
Peter: That seems kinda late
Peter: We need to be producing implementation reports by early spring
Peter: end of year?
fantasai: Jan 15, so people have time to get back from the holidays
RESOLVED: New deadline Jan 15th for test contributions to CR->PR set of
tests (later contributions accepted for interop tests)
<dbaron> Peter: so March 15 for reviews?
Peter: Anything else on this topic?
Bert: what do we do with the snapshots on the test page?
fantasai: keep them, hopefully make a new one by December (to make sure the
index scripts are working)
<glazou> uuuuurgh, 10,000 tests and 173 days that's ~58 tests reviewed per
day, 7 days per week
dbaron: We don't need to necessarily hold very strict review standards
dbaron: In some cases, checking that they pass in a browser they should
pass in and fail in a browser they should fail in is sufficient
glazou: For simple tests, yes, but when you're dealing with e.g. margin
collapsing
dbaron: fwiw I started reviewing some of the tests I was actioned to review
dbaron: and one of my comments was that these 500 tests should be collapsed
into one
dbaron: There are some cases where there are more tests than necessary
glazou: I'm really afraid that this test suite is going to be the gordion
knot of our spec
glazou: I don't want it to be a blocking factor.
glazou: I'm scared
Steve: Does that suggest we put the test suite up and ask the browser vendors
to run it and come back with the tests that they doubt?
dbaron: That's one of the ways we've dealt with other test suites in the past
peter: That's not a bad approach in general. Assume they're good, and if
there's a statistically-high number of complaints
peter: then look into that
fantasai: I think that's fine, but also if there are some tests that have
been manually reviewed we should keep track of that
fantasai: That's valuable work, and we should be tracking it
Peter: Agreed
* glazou that's also 2.4 tests per hour, 7/24 :-)
<bradk> Are there some tests that can be processed automatically, e.g. by
detecting the color red?
glazou: That suggests we have to cancel our general review of all tests
fantasai: That depends on the contributor. For some contributors, I would
not accept their tests unless manually-reviewed
dbaron: Also, even though we aren't going to require manual review, we
should still encourage them to review them
Steve: If we're doing this then we should also have a deadline by which
everyone should have run the tests
Steve: Running them by March 14th won't get us feedback
Peter: We should expect them to be run as soon as we've closed off the tests
for contributions
Glazou: So we should have a Feb 15th deadline for the first run
Glazou: and then we have a chance to fix the tests and run them again by
March 15
Peter: We'll need multiple iterations
<bradk> crowd-sourcing some tests?
Peter: we have a framework for running the tests, right?
fantasai: We have it, but it's still set up as a prototype
fantasai: We'll have to set it up again after the Jan 15th deadline
fantasai: I don't know how to set it up
fantasai: We have to pass it data from the test suite indexing scripts
fantasai: Dom might know how to set it up
fantasai: We can also use it to report which tests have problems
fantasai: There are two things that need to work before we can use the harness
fanatasai: The harness scripts themselves, and also the indexing scripts,
which feed data to the harness
fantasai: If these are working by Jan 15th, it might take maybe a week to
get things set up
fantasai: if not, then I don't know
image-fit and image-position
----------------------------
howcome: We'd like to use it for other things like video
howcome: We have a tentative resolution to combine them into a 'fit' property
howcome: But I don't like that solution
<plinss> http://lists.w3.org/Archives/Public/www-style/2009Sep/0188.html
howcome: I liked the suggestion of content-fit and content-position
howcome: They plan to use it for the whole SVG canvas
Tab: I'm fine with content-fit and content-position, and I think they
should be separate properties
Tab: The only problem is that it looks like a sub-property of 'content'
fantasai: I think content-fit is more confusing, because I could think it
applies to text content
dbaron: I think a property that applies to image can apply to other
image-like things such as video
Steve: if you said picture, then I'd go with it
Steve: they're all pictures
Bert: Picture is not a bad name if we can go with that
howcome: if picture and image are the same, we should go with image since
that's what we use on the web
Tab: If 'content' can create a replaced element when given a url, then it
might make sense
Steve: replacement-fit?
fantasai: doesn't make sense for SVG. Also nobody outside this WG will
have any idea what it means
* fantasai thinks image-fit and image-position are the best names we have
so far
howcome: If it's between image-fit and content-fit I suggest we lean towards
content-fit for reasons of diplomacy
Steve: That's a first
Bert: I don't understand what SVG needs this for
fantasai: mapping preserveAspectRatio to CSS
* alexmog managed to join just in time for the last item on the agenda...
Bert: My impression is that content-fit would mean that anything that is
inserted as part of the 'content' property
Bert: That's not the impression I get. 'image-' would work better
Brad: I agree. When I think of content I think of the content-box
<glazou> replaced-element-fit :)
howcome: We could explain to SVG why we want to keep image-
Steve: Someone explained that the name should be relevant to what people
see, not the underlying architecture
Steve: In that sense image makes more sense than content
RESOLVED: Keep image-fit and image-position as names, explain rationale to SVG
ACTION: fantasai discuss with SVG
* Bert image, graphic, picture, figure, illustration, object...
Meeting closed.
<TabAtkins> Could someone ask Hyatt to come by soon for gradients?
Received on Wednesday, 23 September 2009 22:32:32 UTC