Log file opened at: 2004-03-02 17:07:04
[17:15:44:] <bjoern_h> i would suggest reporting progress since last meeting, even without frenchmen and brains...
[17:15:54:] <bjoern_h> ive completed my action item re libcroco
[17:16:08:] <bjoern_h> no progress on schema, tp and busy and all that
[17:16:39:] <yod> http://lists.w3.org/Archives/Public/public-qa-dev/2004Feb/0002.html
[17:16:40:] <bjoern_h> i will try to get to something next week...
[17:16:45:] <yod> (for the AIs)
[17:17:05:] <niq> yod: AIs?
[17:17:13:] <yod> Action Items
[17:17:14:] <bjoern_h> action items
[17:17:50:] <yod> sorry, I'm just out of a W3C QA WG IG f2f after all, can't blame me for thinking in acronyms
[17:18:03:] <niq> aaargh
[17:18:13:] <yod> there goes the brain again
[17:18:30:] <yod> my two action items are done :
[17:18:47:] <yod> => integrate error messages : went through the archives, it's all in CVS now
[17:19:04:] <yod> => checklink : good progress, still discussing with Ville what the next next steps will be
[17:19:23:] <yod> Ville's Action Item is done
[17:19:54:] <yod> (sent me the changelog for checklink just after the meeting)
[17:20:20:] <bjoern_h> niq, could you summarize our discussion with dodji?
[17:20:25:] * niq tries to remember ...
[17:20:58:] <niq> bjoern was overwhelming dodji with long list of wishes
[17:21:32:] <bjoern_h> bah...
[17:21:44:] * niq thinks we need to prioritise
[17:21:58:] <niq> * Robust parsing - should be capable of recovering - or trying to recover -
[17:21:58:] <niq>   after fatal error.
[17:22:11:] <niq> * Messages should identify where they come from (file/line/col).
[17:22:22:] <niq> * Robust recovery from lesser errors
[17:22:23:] <niq>   e.g. if a value field is malformed, remaining rules in the ruleset
[17:22:23:] <niq>   should still be parsed.
[17:22:45:] <niq> (these are top-prio IMO)
[17:23:19:] <yod> what was Dodji's views on what we'd do with croco?
[17:23:29:] <niq> bjoern was asking for expansion .. e.g. border - > { border-top, border-left, etc }
[17:23:38:] <bjoern_h> re robustness, libcroco does what css2 requires, according to dodji
[17:23:40:] <yod> shorthand extension
[17:24:19:] <bjoern_h> the discussion was more about that this could be done if the information is available through css schema
[17:24:29:] <yod> I see
[17:24:30:] <bjoern_h> that's nothing that libcroco is supposed to do
[17:24:32:] <niq> dodji would like us(?) to work towards xml+css+wysiwyg widget for gnome, if I understood him right ...
[17:24:53:] <bjoern_h> kidding...
[17:25:32:] <bjoern_h> recovering from fatal error is low priority to me
[17:25:49:] * niq would like to generate an O-O repr
[17:25:53:] <bjoern_h> however, it seems that we could extend libcroco to fit our needs quite easily
[17:26:03:] *** Dodji (~Dodji@longchamp-4-81-56-16-184.fbx.proxad.net) has joined the channel
[17:26:05:] <Dodji> hi
[17:26:10:] <Dodji> sorry, I am late
[17:26:10:] <bjoern_h> aha, great
[17:26:21:] <bjoern_h> glad you made it
[17:26:24:] <niq> hiya
[17:26:34:] <bjoern_h> we are summarizing our discussion yesterday
[17:26:51:] <yod> hi Dodji, glad you could join
[17:28:24:] <Dodji> okay
[17:28:30:] * niq hacks together ultraquick outline libcroco wishlist and apologises fro not doing so earlier
[17:28:50:] * Dodji apologises for being late
[17:29:06:] <Dodji> dayjob is sometimes so demanding :-\
[17:29:19:] <yod> no problem, Dodji, really...
[17:30:07:] <bjoern_h> basically, what i want is a new css validator, and to write one i need a css parser. that's where libcroco comes in
[17:30:38:] <niq> bjoern_h: exactly
[17:30:43:] <yod> oh BTW: http://style-sheets.com/validator.asp
[17:31:13:] <bjoern_h> that one never worked for me...
[17:31:46:] <bjoern_h> a { color: stupid }
[17:31:54:] <bjoern_h> result:
[17:31:56:] <yod> I just got aware of this one last week
[17:31:58:] <bjoern_h> "Congratulations! this document validates against W3C's CSS Level 2!"
[17:32:18:] <niq> so perhaps it's just syntax?
[17:33:01:] <bjoern_h> no idea
[17:33:12:] <yod> it's pretending to be clever, didn't have time to check it thoroughly... but we may not want to take too much time discussing this other validator
[17:33:38:] <bjoern_h> agreed
[17:33:41:] <yod> (sorry for starting this digression)
[17:33:47:] <niq> np
[17:34:10:] * niq has done **** all in last few weeks :-(
[17:34:41:] <Dodji> okay
[17:34:41:] <bjoern_h> the wishlist for libcroco is probably quite short, reporting line/col numbers and adding "good" error messages
[17:34:43:] <Dodji> libcroco may not be fully compliants with your needs today, but I agree to spend time on it to improve it for your needs
[17:34:46:] <bjoern_h> that should be doable
[17:34:52:] <Dodji> one should be aware that I do this only during my nights and week ends though. So it will take a little longer than if I was working on it full time.
[17:35:14:] <bjoern_h> well, i do all this in my spare time too...
[17:35:26:] <bjoern_h> so i am used to things taking long time...
[17:35:44:] <bjoern_h> not a problem, i'd say
[17:36:02:] <niq> dodji, we all have that problem.  It means things go slowly sometimes
[17:36:03:] <Dodji> okay. I just wanting to make sure we are on the same track
[17:36:09:] <yod> we are
[17:36:25:] <yod> that's how things for for most of our projects
[17:36:35:] <bjoern_h> (the point is probably, if we dont do it, no one does...)
[17:36:38:] <xover> Regarding error messages: As we've found the hard way with Markup Validator and OpenSP, you don't really want to have to rely on the message strings from a library.
[17:37:07:] <niq> How far does libcroco go in parsing rule values?
[17:37:54:] <xover> (error messages): Depending on context you may want wildly different messages, and kinds of messages (and including i18n).
[17:38:05:] <niq> I found it seemed able to parse things like rgb(1,2,3), but variants on that caused different kinds of problem
[17:38:26:] <Dodji> niq: It tries to stick to the css2 grammar productions
[17:38:40:] * xover is very badly lagged -> apologies...
[17:39:05:] <bjoern_h> if it does SAC, it probably does all we want.
[17:39:26:] <Dodji> niq: properties values are expressions
[17:39:31:] <niq> I think if I introduced a syntax error into something like that it would miss the whole ruleset from there
[17:39:56:] <bjoern_h> that's a fatal error...
[17:40:13:] <niq> like rgb(1, 2, 3)  (with spaces)
[17:40:37:] <bjoern_h> that'd be a bug...
[17:40:55:] * niq would like it to recover better - at worst look for next } and then try to resume
[17:41:08:] <bjoern_h> a too obvious one, so, you probably messed up something else...
[17:41:24:] * niq might have misremembered and fed it something worse than that
[17:41:25:] <bjoern_h> css3-syntax might introduce rules how to do that
[17:41:38:] <bjoern_h> (which is a bad idea, imho...)
[17:41:49:] *** Yves_ (~ylafon@tarantula.inria.fr) has joined the channel
[17:41:53:] * Yves_ waves
[17:41:55:] <bjoern_h> anyway, that's something with less priority imho
[17:41:58:] <Dodji> bjoern_h: yes, that's a fatal error.
[17:42:04:] <bjoern_h> hi yves!
[17:42:26:] <Yves_> hi bjoern, I restarted the validator this morning with your changes
[17:42:30:] <Yves_> (css one)
[17:42:36:] <bjoern_h> Great!
[17:43:18:] <Yves_> I also discovered a bug in the markup validator that way, it doesn't check that a base element may not be after a link element
[17:43:31:] <Dodji> niq: how should a compliant parser behave in that case ?
[17:43:45:] <bjoern_h> ask hixie...
[17:43:48:] <Dodji> niq: I mean in this case of fatal error ?
[17:44:11:] <niq> I'd like to see a "continue parsing" option - like Xerces with XML
[17:44:26:] <bjoern_h> the problem is how to implement that
[17:44:33:] <bjoern_h> its like writing a tag soup parser
[17:44:49:] <Dodji> hmmh, I personnaly don't like that too much
[17:44:53:] <bjoern_h> and starting now to work on it if rules are defined in css3-syntax is waste of resources
[17:45:06:] * bjoern_h agrees with dodji
[17:45:23:] <Dodji> this kind of things made the web be what it is today
[17:45:47:] <bjoern_h> it's useful for a validator though
[17:45:58:] <bjoern_h> but i think this is low priority
[17:46:27:] <bjoern_h> yves, dodji is the author of libcroco, a css processing toolkit in C
[17:46:30:] <Dodji> low priority and hard to code
[17:46:51:] <Yves_> bjoern: yep, olivier told me, thanks for the update
[17:46:52:] <Dodji> Hi Yves_
[17:46:58:] <Yves_> hi dodji
[17:47:52:] <bjoern_h> my question is, where do we go from here?
[17:49:11:] <yod> this is marginal, but if you want collaborative work on a wishlist for a start, there's the new wiki
[17:49:23:] <niq> aha!
[17:49:32:] <Dodji> ah, okay
[17:49:34:] * bjoern_h hates all web fora... but ok...
[17:49:34:] <yod> and I do think a good wishlist edited by the group could be a good first start
[17:49:42:] <bjoern_h> http://esw.w3.org/topic/CssValidator
[17:49:51:] <bjoern_h> for example
[17:49:58:] <yod> yes
[17:51:18:] * niq just posted to list
[17:51:22:] <bjoern_h> http://lists.w3.org/Archives/Public/public-qa-dev/2004Mar/0003.html
[17:51:42:] <yod> thanks niq
[17:52:12:] <niq> sorry, meant to send a better-thought-through list
[17:52:36:] <yod> niq: it's a good start
[17:52:38:] <bjoern_h> my experience with sac so far is that it is good enough for a css validator, it just needs to get the core syntax right
[17:52:54:] <bjoern_h> and since libcroco implements sac, we are halfway there
[17:53:09:] <niq> yep
[17:53:31:] <niq> I'd like to build O-O framework, but that gets much more complex
[17:53:49:] *** Farkas has quit IRC (Read error: 104 (Connection reset by peer))
[17:53:57:] <Dodji> I think I should subscribe to that list then ;p
[17:54:05:] <bjoern_h> i would like dodji to make location (line/col) information available quite soon
[17:54:17:] <bjoern_h> action yod, add dodji to list
[17:54:29:] <bjoern_h> your email dodji?
[17:54:36:] <yod> Dodji, what's your mail address?
[17:54:51:] *** Yves_ has quit IRC ("Leaving")
[17:55:00:] <niq> line/col and file - remember we have @import, and we have <link ...> from *ML
[17:55:09:] <Dodji> bjoern_h: dodji@gnome.org
[17:55:18:] <yod> (Yves too tired, leaving us)
[17:55:20:] <niq> gnome++
[17:55:37:] <bjoern_h> @import should be handled by the implementation, not the sac parser
[17:55:46:] <niq> - nice xml lib, and now nice css lib
[17:56:15:] <yod> ok, Dodji now subscribed to public-qa-dev
[17:56:30:] <Dodji> bjoern_h: okay, for the line/col information made available. I'll start working on it then
[17:56:37:] <Dodji> yod: thank you very much
[17:56:40:] <niq> thanks
[17:57:03:] <bjoern_h> great, the next thing would then be syntax error report
[17:57:12:] <Dodji> niq: some of your requests on the wishlist are more or less fullfilled already, I think
[17:57:13:] <bjoern_h> how would we work on that?
[17:57:19:] <Dodji> niq: the fragment parsing, for example
[17:57:32:] <Dodji> (in the cvs)
[17:57:33:] <niq> excellent
[17:57:38:] <bjoern_h> (SAC is pretty complete for most tasks)
[17:57:42:] <Dodji> one can parse css statements
[17:57:54:] <Dodji> list of declarations etc
[17:58:03:] <bjoern_h> are there interfaces for parse_selector(...) etc?
[17:58:48:] <bjoern_h> CSS::SAC has a parse_simple_selector() but i think thats non-standard
[17:58:53:] <Dodji> yes, you can parse a selector from a buffer
[17:59:03:] <Dodji> cr_selector_parse_from_buf() is the name
[17:59:11:] <Dodji> it returns a CRSelector
[17:59:21:] <bjoern_h> cool
[18:00:17:] <Dodji> really, the weakest part is the error recovering/reporting thing
[18:00:32:] <bjoern_h> yes, how can we improve that?
[18:00:47:] <bjoern_h> (or, how can we help you to improve it :-)
[18:01:22:] <Dodji> today, an error is made of two things:
[18:01:37:] <Dodji> an error message (that sucks) and an error code
[18:02:01:] <Dodji> when the parser encounters an error, an error is pushed on an error stack
[18:02:18:] <Dodji> and also, the error is reported at the sac level
[18:02:29:] *** karlcow (~Karl@ANice-205-2-2-98.w81-50.abo.wanadoo.fr) has joined the channel
[18:02:30:] <niq> can it read error messages as a runtime-thing?
[18:02:53:] <niq> or can it be made to do that reasonably easily?
[18:03:44:] <Dodji> what would be nice, I think, would be to have a structured CRParsingError datastructure
[18:04:17:] <bjoern_h> SAC has a CSSParseException which contains line/column/URI/message string
[18:04:18:] <Dodji> the fields of that datastructure would be things like error code, line/col info, and other things
[18:04:27:] <niq> { line, col, errno, errmsg }, or more than that?
[18:04:27:] <Dodji> bjoern_h: hehe
[18:05:08:] <Dodji> niq: I don't even think that errmsg is a needed. Maybe the translation error code/ error message should be done by the layer above .
[18:05:16:] <bjoern_h> niq, in sac, you would also get the context of the error, due to priour reported events
[18:05:23:] <Dodji> a layer that would be aware of i18n etc
[18:05:26:] * xover adds "byte offset and char offset, from beginning of line and from beginning of file/buffer" to wanted attributes...
[18:05:30:] <bjoern_h> yes, i want error codes, not strings
[18:05:36:] <Dodji> xover: yeah, why not
[18:05:41:] <niq> ok, fairy nuff
[18:05:48:] * bjoern_h would add "original text" too
[18:06:02:] * xover sees snowballing happening‰ÛÏåÊ:-)
[18:06:26:] <Dodji> okay, we all seem to agree on the mecanism then
[18:06:37:] <Dodji> what about the stack of error thing ?
[18:06:40:] <niq> can we rely on original text if something isn't well-formed
[18:06:47:] <Dodji> today, the parser is a simple recursive descent parser
[18:06:57:] <Dodji> oops, soryr
[18:07:09:] <Dodji> niq: yes ? modulo the encoding
[18:07:32:] <niq> the encoding is exactly what I thought might be a problem.  
[18:07:45:] <niq> Did you say it works in utf8 internally?
[18:07:51:] <Dodji> yes
[18:07:58:] <niq> ok, that helps
[18:08:10:] <niq> probably
[18:08:16:] <Dodji> so we don't have the original text really
[18:08:27:] <Dodji> we have it encoded in utf8
[18:08:31:] <Dodji> but we can decode it back
[18:08:35:] <bjoern_h> we do, original *text* != original byte[]
[18:08:45:] <Dodji> bjoern_h: okay then
[18:08:52:] <niq> but near enough, and if we're producing a report as utf-8 we can embed it
[18:09:19:] <xover> Question: Does libcroco have Perl bindings?
[18:09:25:] <bjoern_h> bah...
[18:09:33:] <Dodji> libcroco doesn't have any bindings yet
[18:09:34:] <bjoern_h> i knew that question would arise...
[18:09:40:] <bjoern_h> use SWIG, xover
[18:09:41:] <niq> hehe
[18:09:48:] * xover may be interested in working on that at some point then‰ÛÏ
[18:09:49:] <Dodji> I wrote it in C to easy this kind of things though
[18:10:00:] <Dodji> s/easy/ease
[18:10:17:] <xover> bjoern_h: SWIG is a stopgap, not a solution‰ÛÏ
[18:10:24:] * bjoern_h knows...
[18:10:49:] <bjoern_h> ACTION: xover to write PerlXS for libcroco...
[18:10:52:] <bjoern_h> okay, kidding
[18:10:55:] <niq> xover: is your OpenSp-Perl stuff reasonably complete/stable these days?
[18:11:07:] <bjoern_h> niq, no
[18:11:11:] <bjoern_h> :-)
[18:11:16:] <Dodji> bjoern_h: hehe
[18:11:24:] <xover> :-)
[18:11:34:] <niq> sbp, you going to do the same for python? :-)
[18:11:42:] <Dodji> what is OpenSp ?
[18:11:55:] <yod> SGML parser
[18:11:56:] <niq> SGML parser used for validator
[18:11:59:] <xover> There are Python bindings for libsp.
[18:12:04:] <Dodji> okay
[18:12:05:] <bjoern_h> so, since we agree what we want wrt to error report, what needs to be done to make it happen?
[18:12:19:] <Dodji> I think I heard about it before ...
[18:12:34:] <Dodji> bjoern_h: i'll just put my hands in emacs :p
[18:12:46:] <Dodji> bjoern_h: and let you guys now about the progress
[18:12:53:] <xover> Dodji: http://openjade.sf.net/ (just for reference)
[18:12:55:] <niq> dodji++
[18:12:57:] <yod> sounds excellent
[18:12:57:] * bjoern_h smiles
[18:13:08:] <bjoern_h> yeah. anything we could do to help?
[18:13:32:] <xover> Where do we have the Pizza delivered to? :-)
[18:13:39:] <Dodji> for now, nothing. I'll let you know when it's time for testing ;)
[18:13:44:] <bjoern_h> (without in-depth knowledge of libcroco internals)
[18:13:54:] <bjoern_h> Wow!
[18:14:00:] <niq> xover, ouch; /me not well enough for that
[18:14:08:] <bjoern_h> ok, that problem seems solved
[18:14:27:] * yod can have a pizza delivered here in Cannes, you'd have to fly in :)
[18:14:29:] <xover> Dodji: Is libcroco included in GNOME 2.4/2.6 release?
[18:14:32:] <Dodji> frankly, this kind of work is important for me. libcroco has to be a good basic block in my software stack
[18:14:38:] <bjoern_h> thanks very much dodji, your help is greatly appreciated
[18:14:52:] <Dodji> xover: not yet, because I don't want to ensure abi/api stability now
[18:15:07:] <Dodji> xover: but it is used by librsvg which is shipped in the platform ...
[18:15:08:] <xover> Any target for inclusion, or is it "when ready"?
[18:15:22:] <xover> Ah, right.
[18:15:35:] <Dodji> xover: when it's ready :p but my GNOME mates are pushing hard for me to come up with something api/abi stable
[18:15:45:] <Dodji> quickly
[18:15:58:] * xover recalls some issues with OOo and libcroco via librsvg with Fedora Core 2...
[18:16:16:] <Dodji> yeah, should be fixed now :p
[18:16:21:] * bjoern_h would like to switch topic to "testing" now...
[18:16:25:] <bjoern_h> times running out
[18:16:29:] <bjoern_h> and i am tired
[18:16:59:] <bjoern_h> and i need to go -> town and wash my towels, etc.
[18:17:04:] <bjoern_h> and go shopping...
[18:17:11:] <bjoern_h> and i am hungry...
[18:17:15:] <bjoern_h> so, testing?
[18:17:21:] <niq> fine
[18:17:38:] <niq> you've assembled some cases ...
[18:17:40:] <bjoern_h> yod?
[18:17:42:] <yod> yes
[18:17:56:] <bjoern_h> where are we wrt to testing?
[18:18:23:] <bjoern_h> i would like to have something real soon know to make progress with markup validator
[18:18:24:] <yod> not very far, though I'm multiplying contacts of people wanting to develop test management systems and test harness
[18:18:39:] <bjoern_h> s/know/now
[18:18:46:] <yod> and today, in my meeting, we had extensive presentations and discussions on the topic
[18:19:25:] <niq> You mean W3/officialcom want it?
[18:19:27:] <yod> one thing that the experts were saying is that test cases are pretty useless if the collection is not managed, versioned, etc
[18:19:30:] <bjoern_h> i want something that i can run from command line before a cvs commit to make sure i did not mess it up
[18:19:45:] <yod> so we're looking at :
[18:20:06:] <yod> * a simple mechanism to contribute (and describe) test cases
[18:20:28:] <yod> * a simple way to merge these in and manage the collection
[18:20:33:] <bjoern_h> ( http://www.w3.org/mid/B99F8771-5B6A-11D8-998D-000393A63FC8@w3.org yod on testing )
[18:20:44:] <yod> * one or several systems to run the tests
[18:20:58:] <yod> (covering all or part of the use cases)
[18:21:24:] <bjoern_h> we have this human resources problem...
[18:21:39:] <yod> HR problem for...?
[18:21:45:] <niq> indeed
[18:21:53:] <bjoern_h> everything?
[18:21:59:] <yod> right
[18:22:00:] <yod> right
[18:23:01:] * niq has hihg-priority (i.e. paying) pressures for the first time for far too long ...
[18:23:07:] <yod> the good news then is that I am trying to shift my effort to work a lot more on testing
[18:23:31:] <yod> so HR may be partly solved if I manage to push this far enough
[18:23:40:] <yod> and gather W3C people to help
[18:24:03:] <yod> (QA and WCAG want to work on this, I expect them to help if we/I lead such an effort)
[18:24:37:] <niq> WCAG?  Aha!
[18:24:52:] <yod> given what I have learned today, I have a clearer idea of how to manage such a thing
[18:25:20:] <Dodji> (sorry, could I ask something ?)
[18:25:29:] <bjoern_h> sure
[18:25:31:] <yod> niq: yes, not exactly the same constraints as us in their test needs, but close enough that they may want to work with us
[18:25:43:] * yod notes: my network will be cut within 10 minutes... :(
[18:25:47:] <Dodji> wouldn't it be good is some of you could subscribe to the libcroco list ?
[18:25:50:] <yod> the admin just told me
[18:25:56:] <niq> yod: are the SWAD-e(?) folks involved ?
[18:26:03:] * bjoern_h subscribed to libcroco
[18:26:08:] <Dodji> okay, great
[18:26:30:] <yod> niq : I could talk a couple of them into getting involved
[18:26:52:] <niq> chaals maybe?
[18:26:54:] <yod> the risk then might be to have too many people and spend more time discussing than doing actual work, but it's an idea
[18:27:01:] <yod> yes, I was thinking of chaals
[18:27:11:] <niq> yeah.  Hmmm
[18:28:07:] <bjoern_h> dodji, in fact, since you announced libcroco on www-style...
[18:28:36:] <Dodji> oh, okay
[18:28:53:] <bjoern_h> re testing, i guess i wont work on the code as long as there is no test suite
[18:29:00:] <yod> I think the immediate next step in this would be to discuss the topic on the wiki
[18:29:14:] <yod> unless someone is allergic to it
[18:29:21:] <xover> WikiDiscuss--
[18:29:24:] * bjoern_h is
[18:29:30:] <niq> I *might* draft some ideas, along the lines of the cataloguemanager thing
[18:29:32:] <yod> hmmkay
[18:29:37:] <yod> mailing-list it is then
[18:29:47:] <yod> fine with me
[18:30:03:] * bjoern_h would prefer NNTP but... :-)
[18:30:04:] <yod> I plan on editing/refining my thoughts sometime this week
[18:30:24:] * yod hasn't been on usenet for aeons
[18:30:32:] <niq> wiki: willing to try
[18:30:44:] <bjoern_h> i am not very concerned about design issues, i just want it work
[18:31:10:] <yod> bjoern_h : understood, but I do worry about the design issues
[18:31:23:] <yod> a little, at least
[18:31:27:] <niq> BTW, did we ever discuss cataloguemanager in this forum?
[18:31:36:] <yod> niq: I don't think we did
[18:32:02:] <niq> hmmm
[18:32:10:] <yod> but now plan on giving it a look at least
[18:32:15:] <niq> ISTR someone didn't like it
[Log file closed at: 2004-03-02 18:36:32