ER/AU face2face meeting 11-12 May 2000

Amsterdam, The Netherlands. Hosted by CWI. Agenda.

Summary of action items and resolutions

  1. Action: GR look for academic writing standards. take to EO.
  2. Action:CR write a proposal for 12.3
  3. Action:MC propose definitions of blocks
  4. Action:CMN ask the internationalization interest group about readability indicies in other languages, e.g. Fog Index or style/grammar checkers.
  5. Action:GR ask people who do mathematics on the web to find out combinations of elements that would trigger that this is a math equation or logic expression.
  6. Action:DD continue exploring OCR field that can do anything to recognize that there is text.
  7. Action:WC add HTML MAP technique for grouping links to AERT.
  8. Action:CMN send pointer to where his RDF work is at.
  9. Action:WC add Daniel PICS WCAG conformance scheme to Errata for WCAG.
  10. Action:CS ask the necessary folks at Microsoft if the plug-ins could be made available in more obvious space.
  11. Action:WC ask Dick Brown about accessibility wizard.
  12. Action:WC link to MS powertoys (from list of existing tools).
  13. Action:MC make the browser support info from bobby available.
  14. Action:?? figure out how to present that info in a usable way.
  15. Action:CMN talk to Daniel Veillard about using RPM architecture to track tools.
  16. Resolved: monthly meeting (agenda created by LK, CMN, WC, and Jutta), investigate tool repository.
  17. Resolved: synchronizing the 2 documents (AERT and ATAG techniques) is not resolved, will take to the lists.stripping AERT to just algorithms keep all of the text examples in ATAG techniques?
  18. Action:LK approach Jon Gunderson about UA and ER working together.
  19. Resolved: ER will approach UA about liaison.
  20. Action:WC and LK update charter based on discussions from past couple days.
  21. Action:GR send e-mail to CS about UAs not showing text of client-side image maps even if alt is defined.


Day 1 - Thursday


LK what you are doing, who you are, and where you are from. I'm at the institute on disabilities. I am co-chair. At Temple University where I work on Web accessibility. I'm their engineering dept.

MK I joined AU in march, I was a consultant. I have joined Allaire officially to work on client-side tools.

GR Here representing VINYUG. I'm on a heap of WAI working groups.

CMN W3C staff work on WAI AU guidelines as co-editor and staff contact. playing with Dan Brickley with using RDF stuff for conformance testing.

Mattias from Adobe. I deal with user interface.

DD W3C. The WAI Technical manager. Protocols and formats - they review w3c stuff for accessibility concerns. Work from France. Organizing logistics so if you have a problem talk with me . I have a bunch of quicktips cards - english, dutch, and other european language.

PB from usable net. started italian now american. online services for usability and accessibility.

GB University of ?? research on usability of web sites. did work on usable net.

JR ATRC University of Toronto. Co-editor on authoring tools. do various educational software.

CR also at U of T. Software developer of a-prompt. editor on AERT.

CS from Microsoft lead developer on home page. also working with WCAG. making msn web sites more accessible.

MC cast in Boston. project manager for Bobby. main contribution to WAI is through ER, but CAST has involvement in other groups as well. overall goals: universal design for learning. accessibility is a pillar of philosophy. bobby promotes that.

WL Smith-Kettlewell Institute. on 5 working group.


LK next topic: general strategizing and planning. what want to get out of the meeting and where we're headed.

Overview of agenda: strategizing, then talk about high level requirements - what types of automatic checks can we perform, then low vision issues - color. at some point we'll want to hear results from chris' color study.

/* charles takes over minutes - not yet available due to technical difficulties */


Mattias - clear it's an algorithm. prototypes not needed.

CMN to show it can be done, to produce code that can be reused (save costs), way to bootstrap getting functionality into standard tools,

prototyping - helpful for us to figure out what should go into AERT.

Expert system - if includes knowledge base then testers won't have to see it for all combos?? good thing or bad thing? if good, easy to use...good. but could it miss things? be used like bobby?

WCAG - minimal requirements for browsers? (follow links, display text)

WCAG b/c of techniques document - perhaps coordinated between the groups? also useful to video tape because many people will miss things if not accessible...have to design the tests well. alt-text survey.

charter - Possible deliverables

charter - participation

/* after lunch */

LK Divide large blocks of info into more manageable groups where natural and appropriate

WL not only that, also is it of use?

MC what is a block? a list may be a collection of smaller blocks.

CR what's manageable?

CS what's natural and appropriate?

MK how are you going to use the page?

CS could alert to blocks that are large.

WC x number of menu items, break them up.

CMN mobile work. in terms of controls, you can count them. used to have something that only took 512 bytes, a little harsh. a spell-checker that works as you go that says, "this is getting long" then you can pick a size.

CS what's a block?

MC borrow from HTML that defines block level elements.

JR if an online book.

DD instead of large, idea is structure. what is the priority. take into account priority.

CMN this is a big paragraph - do you want to break it up? (by seeing lots of words in a "block.")

LK checking for headers?

GB usable net checks for so at least one header per page. flags pages that don't contain any pages.

CS if 1,000 word doc with 1 header would pass.

GB yes.

GR same doc, with lots of HRs, suggest to put headers after each HR. assuming HR creating the break.

CS a slightly different concern. probably easier to find.

LK don't HRs it break up auditorally?

GR no.

LK what about "title" on HR?

GR no. in ie4 - it provides tooltip. ie5 does not.

CMN separate checkpoint for using header elements.

WC reads from WCAG techniques section 4.1.4

CS all are easy to test for. harder thing is long doc, how break up?

JR is this data better to be presented in table.what about a term paper?

MC term paper good question for web. by defn web needs to be structured. need to add structure you normally would not have done.

CMN crude word counting visual way to pick up that "this is a long paragraph."

GR screen readers do this. "this block too long, change font size to get it all on the screen."

DD W3C specs are broken into several documents with a table of contents. some working groups do not provide, "download in one shot." this makes it less accessible b/c you have to stay on line. therefore, they have divided into more manageable groups but less accessible.

MC that violates a different checkpoint.

MK if you have it off line you can not print as one document.

DD there are tools that we use to produce specs. a user agent could do this also. the metadata has not been defined for when to stop following links since you don't want to print the whole web.

GR using IE5 when you choose to save a page for offline viewing. it will ask how many levels do you want to follow.

DD still doesn't work.

GR yes, the open issues list you end up with a lot of the mail archive.

MK yes you get things you don't want and miss things you do want.

GR yes, metadata.

CMN different checkpoint. propose: using a crude counting technique as evaluation.

MC need prototype to determine threshold.

WL variable threshold. which elements you key on.

MK if you look at the length of paragraphs, looking at the variation in length.

CS writing style checker?

GR if you want to put a term paper on the web, there are academic standards for publication of papers. perhaps we ought to approach them and ask, "this is for print media, what about web?"

@@GR look for academic writing standards. take to EO.

@@CR write a proposal for 12.3

@@MC propose definitions of blocks

CMN as a person you can count the # of ideas in a paragraph, not sure if it is feasible in the real world?

CS wondering why i was wondering about tools to check writing styles. Word has a tool for checking reading level, is there one that "do you have a topic sentence?" etc.

CMN word's grammar checker checks some things.

JR word has an "auto-summarizers."

CMN it only works for dilbertese.

LK there are some pages which instead of too few blocks have too many.

CS I have usability data on portal data. the more links, the more people will click on each link. tracking of user behavior.

MC no upper limit?

CS expect we will. not yet.

LK could get a lot because thrashing around.

CS found several counter-intuitive things.

MK could be interesting to see how many links are followed by the back button.

LK can have links on page, but have a visual organization to help you zoom in on something, like in the grocery store - they group like things.

CS or placing the things that people like by what they want you to buy. like in Walmart - the best selling things is bananas, there are bananas next to everything they want you to buy.

LK this rule about how many words in a block...once too many words per block, so many blocks before organization of blocks. is the number of blocks smaller than...ok to have 200 word paragraph but not 200 sentences in a paragraph?????

JR we've been thinking about text. lots of things are considered information. sizes are different for different elements.

HB the usability folks have commented on the 7 +- 2 - 7 is the middle grouping.

MC short-term memory retention.

CS you're not talking about short term memory.

HB for gregory it wouldn't be.

MK if you can go through the list, it's searchable.

HB depending on your language, you may have 26 (more or less) letters to look through.

CS there is a number where that is true, but that changes.

LK if you have a bullet list, nominally it is 7. if sorted numerically or alphabetically, then relax that.

CS sub-hierarchy. 5 major groups, several minor groups. can ignore minor until need them. very design dependent.

MK organized lists: are they key words, or starting letters of titles, or ???

MMP different lengths for different languages? in Czech it is hard to determine where the sentence ends/begins.

CS someone must have done research on this. we can't do this.

CR in AERT we've been doing that by characters.

MK I've been working on localization of homesite (German and French). German is a lot longer. not only translate words, have to reformat the layouts.

JR perhaps Dutch is chunking in a way that English can't

LK in learning disabilities, they must get into things like this

MC not my field of expertise, although it is CASTs. the general rule, the better organized the better it is for

JR who does this benefit?

WC reads from WCAG:

Grouping elements and providing contextual information about the relationships between elements can be useful for all users. Complex relationships between parts of a page may be difficult for people with cognitive disabilities and people with visual disabilities to interpret.

LK what if use SPAN and DIV to divide up a page? they are theoretically better things to do, especially if you turn off style sheets, will they do anything?

CS someone could write a user agent that could render each DIV as a separate page.

LK any guidelines that SPAN and DIV should be identifiable?

MC they are generic elements (DIV - block) and span - inline.

CS they show up in the DOM. special tools could be used to chunk them.

WC UA and WCAG met to discuss DIV and MAP to group links. believe they have suggested to navigate through blocks.

WL conceptually any element can be used as a structural anchor point.

CS anything that shows up in DOM>

WL we can make a prototype and show benefit.

GR show me everything in bold, would be useful.

LK if extend this to repair tools, if there was a page with bold, what would you convert them to?

GR pseudo-elements to mark as important. i've been talking with developers re: aural CSS. play sound during is important. e.g. get to checkbox that is checked. the aural portion is something that happens only when a page as whole is being processed aurally.

CS want to see aural css as part of css.

GR want to see it involve serial flow.


WC only thing it says is "...."

CS fog index in other languages?

CMN in japanese something similar.

HB doesn't do much good.

WL basic english.

CS if core language is for college students...

MC down the road there could be RDF that specifies the audience, then the page could perform FOG index based on that.

WL envision anything from 1,000 basic english word vocabulary. if you are trying to write for incoming high school students....determined by core vocabulary rather than word length.

JR some things don't write simply although they could.

CS contracts.

/* laughter */

CMN Word checks for multiple relative clauses, relatively easily found.

MC if we could agree on a mechanism, then it would be "present index to user"

JR computational linguistics.

CR for every language.

LK an obvious thesis topic.

GR "please pick one."

DD you can be compact if your audience understands your vocabulary. this could be considered not "simplest." a word that many may find complex, since they have to look it up, might be the simplest.

MK and the clearest for that audience.

GR use Ruby to provide alternative words.

WL conceptually we can design something to ferret this info out whether useful or not.

CS sounds like a gigantic project.

WL at the simplest level, use stuff from style and grammar checkers. the concept of using those things to check this are available. they can be done, thus it is ok for us to use.

HB usability issue or accessibility?

WL accessibility is subset of usability. if not usable, not accessible.

CS this issue more than others depends on audience. be careful that all pages are not required to be written at 3rd grade level.

WC idea of semantic degradizers so that not summarizes, but makes simpler, or even converts to images (Kynn's suggestion on WCAG list). these are transformation tools that we could discuss.

JR those are user agent things. a commercial web site would not provide.

WC right, transformation tools or proxies.

CS authors should not be required to provide secondary version.

MC if writing to a technical audience, then that audience should understand.

CMN there are 2 assumptions under that: all the people in that audience are those that haven't been excluded from it already. the checkpoint text says, "use ... appropriate for a sites content." the content will determine the simplicity and clarity of the language you can use.

WC students.

CR if English we will check with Fog index, for other languages..."put up a note."

@@CMN ask the internationalization interest group about readability indices in other languages, e.g. Fog Index or style/grammar checkers.

11.1 - W3C technologies

WC reads what is currently in AERT for 11.1.1

WL nothing interactive in W3C stuff?

CMN can do, but remarkably tedious.

MK if can't do on client-side, it is cheaper and faster. can do it server-side but have to go back and forth all the time.

CS communication to 2ndary server is often not feasible.

LK response time is more than convenience, sometimes you can't accomplish without quick response. for example, keeping information in short-term memory while waiting for response.

CS also an issue for someone with ADD.

/* agreement */

WL how do W3C technologies not include stuff that javascript does.

MC another thing, DHTML, is not shockwave so it is accessible and interactive.

CS it uses the DOM - W3C, but done with Javascript or vbscript.

CMN also, we don't have anything to replace JPG. PNG is meant to replace GIF.

DD 11.1 is so mixed up with guideline 3 (using markup properly). it should be removed. 11.1 should be using w3c technology where there is a w3c technology. proprietary HTML should be with "use valid markup." the issue is with word files. you could ask, "do you need word?" this should be HTML. same with PDF. a lot of cases people use PDF.

WL do we have the techniques for the first things you said?

DD flash is used, we can't say use SVG since it is not supported. however, we could say to use in the future.

WC reads technique for 3.1.1

WL do we have tools to do this?

CS we can recognize that there are gifs...

MK mathematics easier than text.

DD w3c logo would not be understandable by OCR.

LK what about when you put a well-formatted math equation through OCR? can it identify it as an equation? e.g., look for equal sign.

WL every time an image, ask if equation?

GR context: math, physics, there are patterns to identify.

@@GR ask people who do mathematics on the web to find out combinations of elements that would trigger that this is a math equation or logic expression.

WL Kurzweil's book reader.

DD collected page of logos. challenging OCR community to run software to see if they can find the text.

LK logos is a tough thing to do, since logos are meant to be visually distinctive. many pages with images, can't tell it is an image it looks like text. i bet OCR would not have images with those.

CMN by and large it will recognize it is text. combine OCR with dictionary lookup.

DD we're talking about 3.1.1

@@DD continue exploring OCR field that can do anything to recognize that there is text.

WL Kurzwiel reader, Arkenstone.

GR usually then read it through scan tune. <"these clumps of things occur in these other clumps, different occurrences...because of frequency of occurrence i can autocorrect...."

WC reads browsers that support MathML.

LK is this accessible?

WL we can make a text-math sniffer?

GR image, superscript, ... is pattern i'll be testing for.

CS are character entities accessible to screen reader?

LK this is a little off-topic, but if i go to temple and say let's make calculus course accessible, is there anything hat i can tell them to do?

CR there is a conversion program to convert MathML to Nemeth code Braille. MathML is coming along.

DD solution today is to provide an image of the equation but a pointer to the MathML markup.

MC on some of the w3c documents, if we have identified an image that uses math. then support putting math in as object with inner object with image....or link to mathml browser.

GR or verbose statement in english.

WL because of amaya your answer is yes.

CMN amaya is not very accessible.

GR one item, if using PDF use a converter. we need to add something that, if you do that, you must compare textual output with original to make sure it is a true representative.

HB concerned that suggest using PDF converter since it goes back to HTML 2.0.

CMN yes, not sufficient.

DD how do you recognize that there is PDF, xcel, etc at the other end of the link.

CS check mime type. follow the link.

DD but, most will probably be, "file.pdf"

LK if it's foo.asp then anything could be coming from there.

CS but if follow the link, could get mime type.

DD 11.1 needs to be cleaned up. says, "convert quicktime to SMIL" from SMIL use quicktime. so SMIL does not replace quicktime, just provides framework.

CS in cases where we say, "convert gif to text with style sheet" are we providing tools to do that? for example, flash to svg?

DD adobe has a flash to svg converter.

MMP LiveMotion...

LK if it's a moving picture, you can convert to an animated gif or svg with animate. if it's interactive flash then those don't apply. can you look at the flash to see if it is animated? MMP LiveMotion is used to create things. Have choice: save as jpg, flash, or svg. can not open gif and save as svg.

CS not aware of a tool that saves gif as svg.

HB suggest CGM.

LiftOnline demo by Paola

PB different types of sites: e-commerce, corporate, etc. different rules based on which site.

DD where would W3C site fit?

MK I don't see my site: technical documentation/online help.

PB we can add another category.

CS are the rules available?

PB somewhat proprietary, yet on the software.

WL can you affect what you want to look for? do you know the name Jakob Nielsen?

GB yes!!

PB from this report we have info we collect info. today we launched the Web Usability Index. It shows the average problems for each category. We can see that 12% of corporate don't use alt-string.

LK what is the rating? is 1 lower or higher than 2?

GB 4 is highest (catastrophic), 0 is lowest.

PB we average all web sites that have submitted to our tests.

WL over a year's time can you determine an increase in the presence of alt-tags?

GB we started 2 months ago.

HB can you use the e-mail address to send a report to someone on their site?

GB no.

CS assuming your audience is site owners.

MK does it do the whole site?

PB no, need a subscription. ok, this site is rated good. lists problems in categories, no catastrophic...etc. gives details for the report.

CS the page says, "too heavy." how big is that?

PB depends on the device.

GB set a parameter of how long a user could stand waiting for the page to download.

MK depend heavily on the type of site. e.g., an options site people are will wait for 10 minutes.

GB this version, does not come with all of the customization that can be employed. on the final version, yes you could do this.

PB creates a report that is printable.

LK in addition to giving the HTML string, does it point to the visual representation of the page by having a flag on the visual presentation.

PB if you go to our web site, you have a few types of tools. Lift Onsite let's you do lots of analysis on your computer as well as fix problems. Not released yet. we'll list new rules in the free system.

CMN with the full service, can you customize the rules? create new rules?

PB you can change severity. e.g., if download time is not relevant then you can make it lower. since it's independent you can incorporate it with any authoring tool.

LK are there other accessibility checks other than checking for alt-text.

PB there are 17 rules we apply for free.

WL what does portable colors do? do you have something that decides contrast?

MC it probably meets web safe colors.

GB yep. or link colors rather than symbolic names. there are a few rules: should use rgb.

WL in principle, can all of the things we've been talking about, we could build these things in here. if there was a lookup table to look up colors, this framework would allow accessibility checking.

GB yes, if automated. there are some 60 rules we are developing.

WL have you got any ideas from what we've talked about.

GB there are some that are overlapping.

PB there are a number we could incorporate.

DD what using to parse and validate HTML?

PB we are in the process of changing our parser.

DD you can use the HTML validator and HTML parser - it's open source.

CS is the rationale for the rules available?

PB in the description of each problem, that info is there.

DD why don't you want image in link?

PB because if you disable image loading and want to follow the link, you have to click once to load the image and again to follow the link.

DD but you are introducing limitation on user based on user agent limitation.

GB gets an excellent rating.

DD what's the issue with stretch gif?

GB image size is larger than the GIF size.

MK for spacer gif, that's not a problem.

GB we try to determine which images are spacer.

WC other version be ready?

GB September

HB privacy concerns with information collecting?

PB we only use e-mail to send info.

CS you store info about every site that has been evaluated.

GB not publicly available.

PB you can select not to.

LK these advance features that come only with subscription, is there some way for us to get a demo version?

PB yes. demo version of onsite and can get some free reports. the running demo is on the Apple web site. they print it on their web developer cd.

WC how differ from NIST tests?

PB not sure.

CMN different usability rules in different cultures?

PB haven't thought about. obviously, those related to language.

MC selection of color.

CS icon selection.

WC NASA has studied.

HB any reason to adopt the WAI priorities?

JR they are not all WAI things. only 2 are WAI things.

WC they use a 4 for no-alt, i think it should be a 5 - catastrophic.

WL perhaps not for their customers.

GR yes! for their customers.

CS are your checks a superset of the WAI guidelines do you cover all of them?

GB no, but we plan to.

HB you could have one class be WAI specific.

CMN there is a real different between the catastrophe between alt missing on image versus alt missing on major image.

GB we guess which are spacers.

PB the tools that we develop, they improve the usability. we cannot guarantee it will be usable. essentially, if you look at the result, they are improving.

CMN do you go back and track sites?

PB yes. we will publish on the index some information about the sites that have been Lifted and how they improve. which percentage of sites have improved over one week. i see people checking the same page 3 or 4 times. each report is an hour after the other. you see the results improving. you see a smiling face. for the user, if the site is smiling or sad that is important. if i say that your page is 43.5% accessible, it doesn't matter.

LK you might want to look at sites where the e-mail address where the site is returned to the same domain as the site is designed.

PB don't do yahoo or hotmail because may be a spammer.

CS i tried your tool on MSN, it timed out. you may have been blocking me. i was sending it to a hotmail account.

CMN interesting how you assess a site like that.

MK an educational tool. see how sites improve.

PB yes.

CS someone who is testing yahoo may want to see how your tool works who doesn't have on their site.

PB we get a lot of spammers.

GB it takes a long time.

CMN can you cache the results?

DD how do you decide? if it's me correcting my site, then...

CS if a large site wants to subscribe to your service, we could use it.

PB yes.

WL how does your site fair?

PB in between amazon and yahoo. we take care of the major things. i'm sure there are problems.

LK does the bobby server get loaded down with people just having fun?

MC no. but people will test their whole site on the web version - presumably some script that keeps sending pages. but we don't get too many repeat hits.

LK is there a way to foil those?

/* adjourn. meet at the Dam at 7:30 */

Day 2 - Friday

Adobe GoLive demo by Matthias

/* begin day with Adobe GoLive demonstration */

MMP have an inspector. can click a button to turn on/off images. it is not an "automatic check" but you can easily see where alt-text is missing.

/* there is an icon with the alt-text written over it. */

LK does the alt-text wrap?

MMP no.

GR tooltip?

MMP no.

WL how are you notified that there is no alt-text.

MMP just by its absence.

WL what if you can't see.

MMP several features built into find dialog. you can search through regular expressions. can configure to look through specific attributes, etc. Can find by element. Could search IMG elements for alt="". Search/Replace also has several interesting features, like "replace element with its content" in which case we could remove <B>.

HB that list of attributes are not specific to the element you are looking at?

MMP they are not the set of all defined attributes for that element, they are the ones that appear in the document. Can see a table of all elements on the page with their type, information about them, and a comment. For example, the attribute "controls" for IMG is for "explorer only." (explorer only is in the comment column).

GR to get to the attributes, do you have to use the mouse?

MMP can tab to them.

LK with the tag replacement function, someone could global search and replace for "font size=+4" replace that with H1.

MMP should be possible yes.

LK can you store that search and replace?

MMP yes.

LK then a company could create a set of macros for people to change physical tags to logical.

MMP yes. there is a "save task" option in the search/replace feature. then, right in the dialog is a button to list all available tasks. For example we have, change header level, find img without alt, etc.

GR and LK had an interesting example that might be possible, MMP was not sure. something about replacing an element with SPAN and class.

CMN does this handle javascript?

MMP can also search the whole site for images without alt text. i see that so far only one page is ok. this page has 7 that are missing alt-text, this one 10., etc.

WL is this an authoring tool?

MMP yes.

WL if someone wanted to have an authoring tool, they could download this from somewhere.

CMN you'd have to buy it.

GR you can selectively drag and drop. can you do this from the keyboard?

MMP complicated.

GR there are drag and drop functionalities in...

MMP wait, here is a button that gives you a modal dialog. so you can also use the keyboard. as a default you get the list of files in this list.

LK if i create a library of search and replace, i could give you a bunch of canned ones that you could plug into golive?

MMP is possible, yes. it is a proprietary format right now, we'll probably change to XML.

LK this is mac, is there a windows version?

MMP yes.

CMN can you transfer the task files between windows and mac?

MMP not sure.

CMN can you use regular expressions?

MMP yes. here is an example. I can search for element I or B. back to my list of images without alt-text, i can click right on one to open it to work on it.

LK could do a search and replace for images without alt-text and replace them with the same IMG markup and then add and image icon that designates this image is missing alt-text. then people when they look at the page can tell which images are missing alt-text without turning images off.

MMP in the special menu we have XML tree submenu. this is a sample, it is not shipping yet. other things...we have a DOM in GoLive. I have an XML page, there is a plug-in. with javascript. it shows the XML tree of this page. I go back to the inspector. this shows the binding between the XML file and the javascript.

LK this is how to use javascript to edit the site? you have a programmable language that goes beyond the search and replace.

MC it's a way of editing your editor.

MMP we also are using javascript to create some key maps. it's not c++ framework. these plug-ins work in windows.

WL can you invent your own elements, and get a list of elements and attributes, and it will go in the XML file?

MMP yes.

MC can you control how they display in the WYSIWYG view?

MMP not sure.

CMN if i invent an element can i invent style sheets to display them?

MMP probably no. not sure.

HB any way to couple a CSS-based style sheet with it?

MMP don't think so. it's only version 5!

/* laughter */

WL does it have a style sheet editor?

MMP yes. open a window for CSS definitions for the current file. I can define name, font settings, etc.

LK could we write a script that would go through arbitrary sites with "FONT" and define classes, and style sheets and "stylefy" them.

MMP not sure. never done it.

LK we have an evaluation and repair platform that is extensible or more sophisticatedly with javascript.

HB apt to port to windows?

MMP yes. already.

ATRC A-Prompt demo by Chris

CR It's an evaluation and repair tool. It's checks them for accessibility according to WCAG. we're trying to implement the whole AERT document. This is a demo, very much in beta. /* run it on a file */ on the left you see all of the problems, on the right you see the instances of it. you can repair this group of problems or each individual problem.

MK does this look for spacer gifs?

CR there is some logic built into it. if long and narrow will suggest use alt-text for horizontal rule.

MK a spacer should have an empty alt-text.

CR yes. we suggest using " ". there has been a lot of discussion on this.

MK if you put in a space, you will get a pop-up. if you put in empty string that won't happen.

CR you can put in null alt-text put we will flag it.

GR techniques in UA, configuration option is to give the user the option to determine how to display.

CR it is an argument about nothing.

WC literally!

CR we have an alt-text registry. each time it sees alt-text it will save it in a file. if you run into the same image again it will suggest the alt-text it had seen before. we've now repaired one of the images. there is a progress bar that shows how many of the problems you have fixed. if you have a small image it will suggest that you use "bullet" as alt-text. on the bottom of the dialog, there is a "repeat repair" that will change the others.

GR is "bullet" the default?

CR yes.

GR can you have a dialog that suggests "*, bullet", etc.

JR what does AERT says?

GR if someone looks at Lynx-me, they will see * and see that it looks better.

CMN * is a better default?

GR yes.

LK a couple years ago, i did a study, some people liked * some liked "bullet"

WC depends on the context. if looking at a page about guns, "bullet" won't make sense.

GR sometimes would like to hear "item."

JR if that's a better default that should be in AERT.

CR alt-text brings up so many issues.

LK let's focus on the tools and not get into the guidelines issues.

CR in AERT there are about 80 techniques. we've implemented 45 in A-Prompt. we're validating the page and will give you a rating at the end. it will give you a WCAG conformance level at the end. You can choose then to only view the Level A issues with your site. In "conformance level advanced settings" you can also pick and choose checks from other levels.

LK when validating documents, there are manual checks, how do those enter into conformance?

CR we have a manual check for such things as language changes. this means having a series of dictionaries for different languages and if any word is not in any dictionary. we can't do this so bring up the dialog for each document. you then have to check a checkbox that you have done the manual check.

LK that's great.

CR we'll have to see what users think about it. they may not want to do that for every check.

CMN if several manual checks, do you have a dialog for each one?

CR no, we have one dialog. people would scream if they don't.

HB can they check the box without doing the manual checks?

CR yes, there is no way for us to determine if they have done them.

CMN word does the language checks.

CR they are more sophisticated! another features - automatic repairs. the user doesn't have to do anything. for example, blinking or scrolling objects. we can detect it and either remove a blink or marquee elements or change it into something that the user configures. such as bold or italic.

WL can the user turn it off?

WC UA side.

CR but blink and marquee should not be there. non-w3c. another automatic check - is to add default language.

CMN is there a wizard mode?

CR you can pick or choose or you can do a whole bunch in a row. i'm showing a few at a time so we don't have to go through all 27. here's our ascii art description. it will detect if you have ascii art. you should have a description and a link that skips over. you can check if your doc already contains a description of it, or you can write a description. we'll link to it. or you can write it into the file.

MK how do you detect ascii art?

CR described in AERT. basically - if in PRE or XMP, is more than 4 lines, and sequence of 4 similar characters.s

MK that would also include code samples.

CR yes, then you would check "not ascii art."

CMN and should be marked as "CODE" not PRE.

CR i'm going to add a check box that says "convert to code." it also pulls up guitar tablature.

HB this will also miss emoticons.

CR not included in this. this is multi-line ascii art.

WC emoticons included in a checkpoint. technique is to replace with ABBR and title="smiley" or something.

CR can also play audio and write a description file. we will link to it. ... headings should be marked up. if we see a sentence that is all bold. we will flag and say, this should probably be a header. we'll suggest a header level.

HB are you maintaining a hierarchic map of what level it should be?

CR yes.

DD can you go to the document to see if it is really a header.

CR we'll show you the text.

DD can you see that section of the document?

GR but, you could open a browser.

DD and synchronize it.

CR idea is that they dialogs will occur as you are editing so if you just enter the text you will get the dialog. nesting headers properly is still under discussion. currently, there should only be 1 h1 in a document and subsequent headings should follow sequentially.

WL what if 2 h2s and one comes after an h5.

CR that's ok. this is according to AERT. AERT needs work. another repair, image maps. if you are using server-side - it will be flagged. if you have access to the server-side file we can convert it automatically to client-side and ask you to insert alt-text.

DD what about grabbing the title of the page that it links to?

GR if you're on line.

MMP do you also highlight the regions? you have a list of areas and URLs.

CR that was displayed in our old viewer. yes, we plan to do this. anything that IE can display we can display as well. with an image map, you should have text-links as well. you can check "yes I have them already" or "no i don't." if no, then we'll generate automatically.

GR where do they get inserted?

CR we put automatically at the bottom.

GR do you put a "skip to navbar" link?

CR no.

GR when you put redundant text map do you use the MAP element?

CR I think it's a paragraph.

GR if put it as a MAP then ATs will recognize it.

CMN or skip the map element.

GR it's logical because the image map is a map.

@@WC add HTML MAP technique for grouping links to AERT.

LK if have client-side imap wouldn't you want to hide them so you don't hear them twice?

CMN could do with CSS.

GR or could use HR with title before. or title on DIV.

LK is there something else with aural cascading style sheets.

GR only screen reader that supports acss is emacspeak. hiding doesn't always hide it from a screen reader.

LK not hiding from sighted, from AT. although, hiding it from sighted will be important for designers.

GR yes, and then AT vendors will have to be aware of to ignore or read things set as hidden.

CR all of this is described in AERT. therefore, if others want to evaluate, it's all there.

LK the guidelines talk about the algorithms they don't talk about the interface.

CR GB, you do repair in your pro version. your repairs will be different. you are welcome to use these features. the idea is to have consistency so that we should all flag no alt-text. ... now tables. we ask is it layout or data. you have to check one or the other. the headings are marked as header cells, they should be marked using standard formatting.

WC should be moved in with other cell and marked as header.

CR if using a table for layout, people want bold and italic centered text. they will mark as TH.

CMN should use the HTML table header column. the repair that i would suggest is to convert the heading cells into HTML headers and merge the two columns. it will linearize cleanly with good structure.

CR you can do that.

LK right now, if this table is linearized it will say, "all about dogs" then "all about cats" then the text with those.

CR right, we have a tool that steps through the table linearized. /* it shows this reading order that len describes. */ the user will have to fix this.

JR automating that would be hard.

CR to linearize problem you may have to move the cells around. but merging a column, we can do that. /* back to fixing this table */ it is also missing a caption. next example is a data table. data tables should have column or row headers. we will let you convert top column, or a row. also to add abbreviations for long column headers. you may require further markup.

HB conscious of writing direction and internationalization.

CR does not have an effect for linearizing a table. linearizing means how the table appears in source.

CMN but may want to go left to right, as with Hebrew.

CR we could check the language of the document before linearizing it.

CMN dir attribute.

LK but linearization order is not affected.

CMN if going cell by cell, first cell from left to right. if using Hebrew braille program you'll start on the right.

LK linearization is taking the order in the source. would be surprised if it reads other than it is in the source.

MC i know text reads left to right, does a table automatically flip itself?

GR that's what "dir" is for.

CMN there are some arabic browsers that support this.

CR we've now finished the repairs, the progress bar shows we're done, and you are AAA conformant. what would W3C think if we allowed people to include this? what assurance do people have that they are conforming?

WL on honor system.

GR ask them to check the manual check at the end.

CR but they can check that box at any time.

GR but at end, before get symbol.

CR so nag them again just for fun? always on honor system. even if do include alt-text no guarantee it is not gobbledygook.

LK alt-text for WAI conformance logos should be "we claim..."

GR the model for the conformance claim is a link to what it is, who makes it, etc.

LK as part of the tool, you check that when that logo is on the page, you link to the official language to check proper use of the icon.

CR I'm shooting for a fall release with as many of the techniques implemented as possible. we've been talking with vendors to get this incorporated into their tools. perhaps also Amaya.

WC what about open source?

CR there was resistance because of 508 in the U.S. concern that someone may hijack it and sell it as a 508 tool.

LK but there are licenses with open source.

CMN if you use this stuff, then anything you do with it must be public.

CR the way we could do this, is to brand it with the logo. with open source, you would not get the logo. instead get the WAI A-prompt logo. if you produce your own validator, you could use your own logo that says "validated by open source brand x." then you would know that this is different than someone else's a-prompt.

CMN this seems like an odd reason for not making it open source. if concern is that american companies will buy dodgey thing for compliance.

LN this would be so useful. but concerned about costs.

CR thinking about licensing for free. therefore if frontpage wants to use it, they could integrate it.

MK any way to talk about integrating? how would you do that? show a line number in the source code? we would like to integrate it will HomeSite, if have line number then you can show where to go.

CR this is doing evaluation and repair. you can put up your own dialogs.

WL the instance that it finds, it returns where this is in the document?

CR it will know by the code. you can send it either a chunk of code or a whole file. if you want it to send you the line number for errors, we could do that too.

LK how windows-dependent is this?

CR you get back a text-string and a line number. currently only works in windows. talking about integrating it with Bobby.

RDF tools discussion led by Charles

/* charles commences handwaving session on RDF. invisible partner Dan Brickley is not here */

CMN who knows what RDF is?

/* a few hands raise */

CMN its stuff for a computer to read. Danbri and I wrote a schema that says, "we have facts that we can assert. something can confirm to a checkpoint where a checkpoint is a URI. you can take a tool and say that it conforms to these checkpoints. you have a table that lists checkpoints. you can say how they conform. can say, "yes, no, n/a." In ATAG there are some checkpoints with several subcomponents. you can conform at several levels. in the schema you can say, 'you conform to this checkpoint to level 2. the tool reads RDF. you can use RDf to link pieces of RDF together. I have several statements about tools and checkpoints. I can find out what that means about conformance. there are enough statements to verify that Amaya does not conform to ATAG. to make that useful we need something like WART that outputs RDF. It reduces it to info we can suck into the tool. In WCAG we can add metadata to pages to show what they do. There is a PICS scheme written by Daniel that will allow you to make a conformance statement. Search engines can look through this info. This is implemented in JavaScript. There is a prolog interpreter available in javascript. that's how we did the inference stuff. one of the neat things about individual assertions is that i can stop going through the conformance claims.

JR you don't know the priorities of the checkpoints do you?

CMN we wrote statements about priorities. we do have more than numbers. there are assertions about priorities.

LK with these factoids, could you have factoids about the page, then the browser, and then deduce "i can read this page with this browser."

CMN yes you can.

LK are you going to?

CMN simple logical stuff. the next step is to take RDF and suck up info: what is unknown about this tool or this page? if it is known that a conformance happened on date X, and only 1/2 the evaluation was done, can we finish it off? using the WART perhaps. or if you have 2 or 3 conflicting assertions, what can you do with them? if you know who says them, you can say, "I don't trust daniel's assertions, i'll trust charles." in looking at different conformance tests, we hope to do this in ATAG, have several people review a tool. we expect people will come up with different results and want to see where and why. relates to what GB discussed yesterday with tools that do things differently and come up with different results.

LK the w3c takes versions and makes XML out of them. is there an XML prolog in the way?

CMN we've been doing inferencing logic with that particular hack.

LK w3c not writing an inference engine?

CMN no. one thing that danbri thought about (we've been using for Authoring tools) is RPM files - a database of linux packages. there's lots of RDF about them already (who made it, when shipped, etc). We thought about adding conformance info to RPM find info. Then you could say, "find me an editing tool and is AA conformant to ATAG or meets these five checkpoints that I care about."

LK not sure how to express this, you could have an RDF assertion, "checkpoint 3.3 is T or F." could you break that down, where the atomic unit is not the checkpoint but "alt-text" and you have an assertion about alt-text and then make inferences based on that.

CMN yes.this object conforms to these guidelines. this object conforms to this checkpoint (at 1, 2 or ,3 or n/a). if you conform to 1.1, 2.2, 7.8, etc of guidelines x you are level A conformant. in ATAG we have several that are based on WCAG. can say, if you conform to WCAG X,Y,Z you conform with ATAG A,B, C. RDF does not have a way to specify inference rules. if you have the factoids in RDF then you can use any RDF parser.

LK if have factoid for checkpoint 1 and checkpoint 2. but statement: need checkpoint 1 and 2 to be level A compliant is not in RDF.

CMN RDF only makes assertions.

LK classic example, "socrates is man..." how much can express in RDF.

CMN Socrates is man. Man is mortal. in RDF. inference engine to say that Socrates is mortal. it needs glue. the other place thought about using it, in A-prompt. you have a manual check like "alt-text is valid." can use to expose pieces of your data.

LK in terms of problem solving: if you're missing alt-text, you can grab the title and bring it back. is there a way to represent that in RDF?

CMN for example, you have the w3c icon. you can say for URI X, there is a description at Y. you don't know what the assertions mean. you need logic in your tool.

MC are you publishing this schema?

CMN I put it on the web when it looks like garbage. sent it to ER in november. I'll send another pointer.

@@CMN send pointer to where his RDF work is at.

WC different than Schematron schema?

CMN yes.

WL need RDF somewhere?

DD is XML so just need XML-parser.

WL does someone have to do something to a Web site.

DD need a tool to write it and read it. it just expresses statements. it sits there doing nothing. it's markup.

LN you can read it.

DD statement itself could be stuck in a page.

WL but if you haven't done that, CMN can't do anything.

LN but you can add it to someone else's.

CMN it can be stored wherever you want. i can make statements about your work.

WL can't be automatic.

CMN unless have machine that does so. WART could be set up to produce results in RDF, then someone could use that info.

MMP same thing that Apple used a couple years ago?

CMN no.

DD it is content free. you need a schema that says, " iw ill speak about violets" the schema defines the vocabulary. then use that to express statements. started 5 years ago with PICS in Lisp syntax style. you can express a rating system, with labels. mostly used to define level of harmful content (nudity, violence, etc.)

HB use this as a basis for comparing various ER tools? would this be a way to show cover or absence of coverage of various guidelines? is that what you're suggesting.

CMN not suggesting it, but will from now on.

LN it would not be hard to go through a cataloging system to add this info. it's an interesting piece to explore. adding info to the cataloging system.

CMN yes, librarians are very interested in the capabilities. RPM is a catalog and will be using it. e.g., can get the latest version of amaya. plus, there's the "what's related" button in Netscape. it uses the same idea.

HB is it possible to get a meta-level certification of way conformance?

CMN we have that for PICS already.

@@WC add Daniel PICS WCAG conformance scheme to Errata for WCAG.

LN need to register it somewhere so that they know it is another thing to add in.

HB not part of dublin core.

LN it can take extensions. there is a community. it is important that dublin core has an extension for WAI. it would be a wonderful place to add it in.

/* break for lunch */

/* gregory performs interpretive dance to funky music. the two pieces he performs are: blind man walking against an inaccessible panel, and inaccessible. */

WAVE demo by Len

LK In addition automatic checks there are manual checks. the main theme of WAVE is to help people do manual checks. The legend is the summary of what WAVE does. It shows alt-text and a symbol. If blank or null, a different icon (orange with question mark). No alt-text is a red icon, looks like measles on many pages. checks for applets. gives reading order. puts in arrows and numbers for order in which it is read. where it finds logical tags it puts an icon there. when see something that looks like heading, it has icon next to it to confirm. all of the output is designed to be accessible. it is accessible to someone who is blind, but i'm not sure how useful it is to someone who is blind. /* Looking at WAVE results Marjolein's Help for HomeSite Users. */

DD should make icons uniform.

LK wanted something specific to alt-text.

WL what if you click on those links?

LK will not follow the link "WAVE-i-fied."

WL don't address errors in the presentation of the link. like "click here." there are guidelines that a link should do something-a-ruther.

LK this is for visual inspection. if a link says "click here" you see that.

WL what about length?

LK this does a fraction of the guidelines. it does nothing with length or content of link text.

WL don't links have attributes.

CMN open source?

LK want to make a proper perl module that i can put on cpan without blushing before i make it open source.

WL does this summarize the issues.

LK that would be a nice thing to do.

WL it's called Bobby.

LK a non-intersecting set. i'm focusing on manual checks. some of the automatic things are obvious on inspection.

/* going to site with inaccessible features - yahoo */

LK WAVE also does image maps. it does the same alt-text labeling for each area in a map. alt-text of image map says, "yahoo (see areas of image map "m")" have to go find "m." with alt-tags on areas, sometimes a tool tip will pop up and sometimes they won't.

MK putting in a title instead of alt.

WL often display title, then alt if no title.

LK when i talk with people not familiar with web accessibility, if only looking at tooltips to talk about alt-text will only see those that you mouse over. WAVE can be powerful to show the discrepancies between images and alt-text. Great for educating. If you look at a table for layout, visually it makes sense, but if you look at the reading order you get something that doesn't make sense. this helps people understand the order in which things are read and then they can understand the problem. this would be good to give to a manager who can say, "yes, there is something wrong here. go fix it!" or give it to testers. you are not talking in terms of HTML code but the user code. is not as "black box" as testing with a screen reader.

DD a more advanced table linearizer, like tablin, could repeat the header for each cell. therefore that not as much of a problem. it depends on the user agent features what reading order they use.

LK for making that judgement it would be useful to identify which are table headers. that's easy. i'll add that. you still need to know what's going on. the help is not built in. favorite site?

GR a site with flash.

LK dont' handle flash.

WL we need something! that's where we're headed.

CS handle OBJECT and EMBED elements?

LK no. need to show OBJECT and type.


LK this is a measle site /* lots of red images, i.e. no alt-text. not much else */ you can show something that says, "no alt-text" but to see icons over and over. to see your own page splattered with icons seems to hurt. trying to make it as painful to sighted as to person using a screen reader. /* temple home page */ lots of spacers. in a table had two images "temple" and "university" but "temple" in it's own cell and followed in the reading order by two cells with links then "university." so they put "temple university" as alt-text for "temple" then a space for "university."


LK there is an ad for flowers the alt-text is "banner." the advertiser is getting shorted.

WL they've cleaned up their missing alt-text.

LK right. no more measles.

CS do you check for LABEL elements? that will make a difference in reading order.

LK there are things it doesn't display. how much do i put on the display. I am afraid of it becoming too cluttered.

CS you could be giving incorrect information.

LK hmm, that's true. do screen readers honor label?

GR yes.

JR just like TH will change reading order.

LK if read top to bottom, does it change reading order or only when go by element?

GR if form in a table...JFA will let you move to 1st form control and will speak it whether in table or not. if go into forms mode.

Amaya demo by CMN

CMN how many know windows?

/* lots of hands */

CMN how many have used Amaya?

/* all the same hands */

CMN /* looking at a page */ Amaya came out of a project to get people to do structured editing. Can look at this doc in structured view. You can edit in any of the views there are about 5 of them. Content view: basically an outline of H1s and H2s. Links are links can follow. On the W3C site, Amaya uses an HTTP Put to publish. If you have a server that accepts puts you can save right to the web.

CS security?

CMN have to be authentication by the server. Links view and where they go including images.also a text-only view. can look at any attribute. can "show targets."

CMN written in C. open source. can navigate by keyboard. in contents view can see headings and how set up. which are targets. Help is not brilliant, but we've been writing up how to use it accessibly and "how to write HTML." HTML is a language to markup semantics. no way to put in null alt, has to be a space.

MK can you work on multiple pages?

CMN I have multiple pages open. won't work across pages. has a make book function can convert into a single book.

GR with the style thing does that put it in the head or can you say external. if i had a class on one could i save it to an external style sheet.

CMN if create a rule saves in head.

WC how work with Amaya to incorporate AERT ideas?

CMN it's in C.

DD worse than C. using abstract UI layer.

CR not peered with OS?

CMN no.

LK with other tools there are higher order constructs that work at higher level than C code. anything analogous to that?

CMN at the moment the only language to that is the transformation language.

LK what about the transformation language?

CMN it's a special purpose language for finding elements and changing them. similar to XSLT but not XML and peculiar to Amaya. Amaya exists as a testbed, so it supports MathML. e.g., how might we implement longdesc?

WC are they doing that?

CMN hopefully in a couple weeks, i'll be visiting.

LK if you make a change and you want to see it in the source, you have to make it available in all of the language translations?

CMN a nice place to test things, great to play with authoring ideas. it is very different from other authoring tools. it doesn't have the same ease of producing stuff that is WYSIWYG. in terms of having a commercial impact it has a lot of catching up to do.

MS Authoring tools plug-ins demo by Cynthia

CS there are a bunch of power toys for IE. there is one called "toggle images". magnification, can highlight a section of HTML and view the source. all MS authoring tools use COM (Component Object Model) add-ins. Any functionality you want to write to check for things, can write a COM tool. Visual InterDev has good alt-text features. in property sheet for images, have property for alt. Because of the event model, theoretically write plug-in that would reacte to the dialog. if view source in word, it opens visual interdev.

LK with the COM interface, it gives you a hook into the user interface. like, when you bring up an image, it could pop up a box that asks for longdesc.

CS yes. there is a whole tree of documentation about the object model.

LK also means that any procedure that you could do manually, you could write a script for. it is operating at the semantic level.

GR the power toys have some neat things, but are not easy for people to find. could they be made available in a more obvious space?

CS yes, i'll take that back to the necessary people. Word and FrontPage support plug-ins.

@@CS ask the necessary folks at Microsoft if the plug-ins could be made available in more obvious space.

WC what about the accessibility wizard?

CS ask Dick.

@@WC ask Dick Brown about accessibility wizard.

CS if anyone is comfortable with COM, let us know.

CMN if we write a javascript feature we can plug it into frontpage via COM. then drop into Mozilla through javascript toys. and put it into golive.

LK COM is not just microsoft?

CS no. it's been around for a while. can write in any language. pretty clean.

CMN Mozilla uses XPCOM, it's just about the same.

@@WC link to MS powertoys (from list of existing tools).

HomeSite demo by Marjolein

MK pretty much everything is configurable (toolbars, resources, editing area, etc.). there are different modes of working. it is designed to help you edit tag-based languages, but if you want to type plain text you can do that too. in an elements soon as i type a space it will drop down all of the possible attributes for that element. i can select an attribute, then pop up another box that lets you select the attribute value (e.g. for bgcolor on body, gives you a color picker). this is called "tag insight."

another way is through the "tag editor." with body and bgcolor, get a preview of colors and body. complicated and can't be described in VTML (visual tools markup language).

HomeSite is completely scriptable. Has an OM. Can create script to install program. Can go to registry and changes toggles. Can set it up, so that it uses VTML instead of regular interface. now, if i edit that tag, i get a different tag editor than before. this one is written in VTML. for every tag in HTML 4 there is a tag defn in VTML. that defines a dialog like this, most are tabbed interfaces. you will find a tab called "style sheets/accessibility."

Since this is written in VTML it can be changed very easily. Any changes will have to release to new version of the program.

MMP do you have a visual editor for VTML.

MK all in same editor. I will show you. In the resource area, i find body.vtml and it is in the editor.

MMP but is there a visual editor?

MK there are tag editors to edit it. i get the pop-ups for the properties of each element.

MMP how do you localize your interface? extract the strings out of VTML?

MK the localizers edit in place and replace strings in source code. it does not have a mechanism to pull in external files.

MMP do you have a mechanism to display the dialog?

MK when homesite loads, it caches the VTML. if you change anything there, you can clean cache and view change. another interface called "tag chooser" shows a few types of tags: browser specific, obsolete, deprecated, general. helps to check the layout of the tag editors. not an official editor, but easy to work with.

CMN possible for a user to control the combinations that are available for the tag editor.

MK also controlled by VTML. could add section for various languages.

CMN validating?

MK yes. a spell checker (on while type), link checker, html checker - validate against VTML definitions.

CS if you delete the alt="" in your example, will it trigger something in the validator?

GR one of the nice things about the validator, it's smart enough to replace character entities with elements. if click on it, will

MK since so open, many people write scripts for the interface and makes them accessible.

/* break */

Bobby demo by Michael

MC showing beta of version 3.2 of client-side version. bobby approved means a page meets all p1 checkpoints including manual checks. 3 columns, 2 numbers each. first is automatically detected errors for that priority level, 2nd are manual. then the browser errors. the browser compatibility checks are run but not linked to the accessibility checks.

DD my question was about the maintenance of browser checks.

MC we have an SGML DTD. problem is finding out from vendors what they support. we are sometimes corrected by users.

GR is that info you could share with people for the browser support page?

MC just post the info? yeah, i could do that. perhaps not in a format that people could understand.

@@MC make the browser support info from bobby available.

@@?? figure out how to present that info in a usable way.

CS give an example of a browser error?

MC purely syntax.

CS if have labels set up, browser not supported, that's not included.

MC used to be in there. no longer. that is hard to maintain. data is still there.

basics: accessibility our primary focus over last couple of years. couple of types of manual checks: some are triggered by element others triggered all the time. several may be triggered for several elements, but we only bug you once. working on a report filtering option. i.e., "I don't want to be bugged by certain things." e.g., only tell me about p1. also, can select "full support, partial, etc" which means only tell me about errors that bobby can automatically detect.

GR want to say you mean that, "full support" not obvious.

CS I thought you meant supported by guidelines.

MC can sort the table by priority, guideline title, select regions. when select a filtering option, report totals change. needs to be clarified. working on localizing on the fly. Will be accommodating other styles in the future.

HB i had tried using scripts to extract info from reports, could not.

MC can export report to delimited file.

Tablin demo by Daniel

DD online interface to a table linearizer. a java program, generates HTML. the tables are linearized. /* shows what it does for */ in this example, the "headers" are not real headers, thus the header and content do not linearize well. if had been marked up well it would linearize well.

WC wonders if we should suggest that part of marking up layout tables with structure is to use TH as well as Hx for headers. we currently suggest, "TD" and "Hx".

CS who is the primary audience? end user or page author?

DD this is a transform tool. it doesn't repair or evaluate, therefore for end users. perhaps attach a script to your user agent to linearize content for you. OR use it as a proxy.

CS I've had people tell me to use it to check pages for how well they linearized.

DD no.

WL but an authoring tool could use this to help satisfy guideline 7 of ATAG [make the editing tool interface accessible].

DD no, it's a user agent type of tool.

CS is it possible to integrate with a proxy filter?

DD yep. it's just a filter. it is a Java servlet.

Communication mechanisms between AU and ER

WL since we're both in the same groups, then the flow is from left brain to right.

CMN how many people in both groups?

/* 2 hands */

WL why does AU still exist?

GR makes sense for ERIG to merge with AU. AU is working on techniques. AERT is a techniques doc based on WCAG.

CS but not for authoring tool.

GR there is cross-over. perhaps a tools interest group that covers all of the groups (AU, UA, WCAG). there should be no distinction between editor and browser.

CMN in order to get from here to there, how does the idea of regularly joined meetings sound?

/* lots of agreement */

CMN what form does that take? monthly? what would be talk about? how would that work?

/* agreement for monthlY */

CMN AU needs to watch AERT, could be item each time. AU has a dependency, say "go read AERT." also, there are the tools pages. AU has a page of authoring tools. ER has existing tools.

HB what about EO?

WC yes, they have a list of UAs.

CMN perhaps give to them?

WC but these are tools we are working on, therefore an inefficiency. excited about the level of configurability and scripting / plug-ins that we could develop. point to tools or submit tools to Allaire site of scripts for HomeSite.

CS several sites like that exist.

CMN how many people have used RPM Find? a giant database of linux packages.

LK general search engine or do the tools use metadata somewhere?

CMN you have to put a package somewhere. the package has to be on line somewhere. then send to rpm find the meta data.

WC web interface?

@@CMN talk to Daniel Veillard about using RPM architecture to track tools.

JR can we develop technology specific things?

/* yes */

MC developing one for dreamweaver as a prototyping tool that will submit to the group.

WL what about the review aspect?

CMN need disclaimers.

WL what about reviewing tools - not all tools should be there.

CMN a couple levels - we can track those that we review and those that are submitted.

CS 2 lists.

DD not excited about the "prompting" part of ERT. ATAG techniques could be prompting and AERT be just review.

WL not talking about the document.

GR you want to see AERT turn into algorithms and triggering mechanisms?

DD yes.

WC does everyone agree with that?

CS no. there is some separation between authoring and repair.

DD confusing because we have AU and ER doing the same thing.

LK post that to the list. I happen to agree with you.

CMN post it on both lists. we're interested in how to build the techniques document.

JR you would suggest for a repair: this is the algorithm to detect, this is the algorithm to repair, no text.

DD right. it's cultural.

JR then should it be in AU? or not text at all?

DD it's more an example flavor. it gives you an illustration. it's fine there.

JR AU might say, "use these algorithms with this text below."

CMN we may also get sample implementations for a tool. the value of having an existing tool that does it is good.

GR not only illustrative but provies it is possible.

WL will help sell the tools.

resolved: monthly meeting (agenda created by LK, CMN, WC, and Jutta), investigate tool repository.

resolved: synchronizing the 2 documents (AERT and ATAG techniques) is not resolved, will take to the lists.stripping AERT to just algorithms keep all of the text examples in ATAG techniques?

WC what about UA? gets back to original set up where WG meets twice a month and IG twice a month. i.e. ERWG meets one week, then AU/ER next, then ERWG, then UA/ER.

LK or what about all three together?

WC yes, gets at the idea of bringing authoring and user agent together.

JR does AU and UA have enough to talk about once a month?

GR yes, because of the partial implementations. ER is the bridge between AU and UA and the user.

CMN suggest that we start out look at having jointly meeting between AU and ER. propose the first meeting of each month since that is easy to keep track of. what time? AU meeting 2:30 on Tuesdays.

HB in the throws of getting UA out the door will probably not be ready now.

CMN yes, busy to get that finished, however have lots of questions that are probably best answered by ER. encourage ER to talk with them.

WC take to CG?

CMN first take to ER list.

LK the only difference to me is that AU seems to have a reference to ER and UA does not. it would be equally useful in both.

WC but IJ didn't see it because nothing in there right now.

@@LK approach Jon Gunderson about UA and ER working together.

resolved: ER will approach UA about liaison.

@@WC and LK update charter based on discussions from past couple days.

LK can imagine ERT having techniques stripped to algorithm and then say, "apply to UAthis way or AU this way."

GR with RDF the user could say, " I just want to see X or Y."

HB if you turn off images you can't depend on anything in the image. if you turn of colors... there are series of assumption.

GR need of repair tools b/c most browsers today don't show text from client-side image maps.

@@GR send e-mail to CS about UAs not showing text of client-side image maps even if alt is defined.

/* continued discussion about using markup - not minuted */

Planning questions

$Date: 2000/11/08 08:17:43 $ Wendy Chisholm