Minutes from 07 November 2000 join ER and AU working group telecon

Available: http://slow3.w3.org/WAI/ER/2000/11/07-minutes.html

07 November 2000 join ER and AU working group telecon

Participants
·       Chris Ridpath
·       Jan Richards
·       Len Kasday
·       Harvey Bingham
·       Mattias Mueller
·       Wendy Chisholm
·       Brian Matheny
·       Charles McCathieNevile
·       Phil Jenkins
·       William Loughborough

Machine readable output for evaluation tools
LK If something needs a manual check so that the evaluation tool does not 
keep bringing up the check, we would like to store state on evaluation 
checks. I'm interested in AU response to the ER proposals. Bunch of 
proposals. You must represent the document in some way. First decision, do 
you mix in the info into the document or do you put in separate file. If we 
want to point into the document how do we do that? How do we represent the 
document? One extreme is representing as a text stream, other extreme you 
represent it as a parsed document giving parse tree. If you represent that 
in XML you can point into it with XPointer.
JR Will this info be stored locally or appended at the end of the file?
LK How would you like it? We've discussed both possibilities.
CMN Look at the upcoming Amaya release that will introduce the annotations 
service which stores outlying annotations in RDF. Would be useful model. 
It's the same type of service as a PICS rating.
LK Does the annotations service, it can point to a piece of the document?
CMN Either piece or whole document.
LK What if it is a css file? not xml at all?
CMN Then need way to point into a piece of it. One possibility, is stuff 
that you can tidy into XHTML. When you get into a file that will not become 
XML, you need to solve that problem at a piecemeal level. That will vary 
depending on type. Pointing into Flash will be different than pointing into 
SVG. Not a fan of pointing into bytecode offsets.
HB All of them are potentially screwed up by editing to the file.
LK Let's say we have a .css file. If we don't point into it as a byte 
offset, what would we do? If not byte offset we're parsing it.
JR Don't know to what extent it should be done, but that system is better 
than byte offset. How much editing can happen?
CMN If pointing into XML, can use a fragment mechanism to identify 
fragments. For integrity checking you do a checksum on the fragment. Then 
you can use this as an annotation.
LK If we parse down to point of a name of a tag then we are in trouble. 
There can be a bunch of "P" elements. Need a bigger fragment? This has all 
of the earmarks of a classic problem.
/* CMN drops off to catch a plane */
PJ When I read the title for the agenda. I was coming from a diff paradigm. 
I have 2 requirements that this XML representation be very scalable so I 
can look at running tools on lots of pages at a time. We're not just 
annotating a page. I want to report on thousands or millions of pages. I 
want to programmatically gather the list of problems and summarize them.
LK In addition to handle thousands of pages, you also want to zero in on a 
particular alt attribute on one of those pages?
PJ Yes, or I may want to say "there are lots of alt-text missing." I would 
not call that machine readable, I would call that human readable. I thought 
of loading databases with results. More of an application rather than a 
document.
LK We were talking about this in the context of, e.g. 2 tools that check a 
web page. you want to merge the results of the 2 tools to figure out what's 
going on.
PJ Can't we have a DTD for people to follow. An XML language for 
evaluations. Most today keep track via line number.
LK I as a human could say, "look at this javascript." would we want to have 
a way for XML to point into javascript? to say "this statement causes a 
problem."
PJ Perhaps. Or say, "here's the the line number." Don't know how else to do 
it. It doesn't point back to the page, it just says, "here's the result."
LK So you want to treat it as a black box? The tool would say, " if i go to 
this page and click on this link and enter these values in a form i get 
these errors."
PJ Right.
LK Like systems tests.
PJ You want the keyboard script for how you got to that point.
HB More complicated b/c dynamically generated page it could change each time.
PJ But it's repeatable b/c you have the steps you went through to change it.
HB Or changing under you.
PJ Like the results from a search. But the characteristics are similar. An 
XML lang for results files is useful. I could take any tool I'm using and 
merge the XML files or pick and choose from different files to create a new 
report.
LK In terms of high-level requirements it needs to point into HTML, specify 
test scripts,
PJ error counts, priority, etc. Like what you get from Bobby, error ID, 
links to explanations.
LK If you have a file which gives individual errors, and leaves the 
statistics to the customer of the file.
PJ If I have the summary data, I'll want to keep that. Why would a machine 
be reading the insert reports?
LK Say you have an authoring tool, you get a report, it flags lines of 
code. If a WYSIWYG editor, the editor could flag issues as you edit. You 
can fix inline.
PJ You want to create interchangeability so that SSB could read Bobby and 
vice versa?
LK Yes. Plus, an evaluation tool may want to read it since lots of human 
checks. A typical tool will ask "is this alt-text correct?" If you say, 
"yes" you don't want that to pop-up again.
PJ So you want history?
LK Yes.
BM Means you have to determine when source has changed so can ask again.
PJ Other people on call that will do this, other than WAVE, SSB, and Bobby?
CR A-prompt.
/* discussion that at least these tools will. how much participation from 
the developers of these tools */
WC Sounds a lot like RDF.
/* WC describes Amaya annotation, answers as many questions as can about 
it. Bring together several W3C technologies as we haven't done before, that 
i'm aware of. */
LK Getting back to PJ's comment about scripts.
PJ There are webcrawlers.
LK Do they insert id's?
PJ Internal ones would. That's how I create a lot of my reports.
LK You have robots that will crawl the site and affect the pages?
PJ I have not filled in forms, but we have clicked buttons to expand content.
LK If there a script that does that...
PJ Perhaps separate the point in the process at which you do certain things.
LK Another argument for summary stuff is that the tool doing the details. 
Much easier for the tool to summarize rather than something else.
PJ What if i merge 2 tools, one found 3 instances and the other found 6. is 
that 9 instances or 6 or 3?
LK When you specify a way to crawl, how do you denote the process?
PJ they just give me a URLs to process. URLs are locally saved pages.
LK INteresting to know the steps they are going to do.
Action PJ: Find out if you can share information about how the instructions 
for "black box testing spiders" are documented. If can share, please do so 
with ER and AU.
PJ Do we need additional info in our scheme. How do you merge data from 
same tools?
HB Same tools running on different docs?
PJ You run site X, it runs 10 processes, I'm analyzing 10 pages at a time 
then merging these results into one report. I may want to do it on the fly 
or after the fact. Just thinking out loud. Not sure if these are additional 
requirements.
HB Just store them all to a database and summarize.
LK May not implement it, but allow for it.
PJ right.
LK We should try not to limit the notation. That's treating the web site as 
a black box. You can have .asp which can have perl, or VB in it. Do we want 
to say things about particular statements in these scripts?
PJ yes. e.g., we have JavaScripts that have invisible divs, it shows up as 
a pull down. We could annotate it to make it more accessible to a screen 
reader. Like WC described, file could be glued together in my browser. 
Pulling info from various places. It's not only machine-readable but human 
readable.
WC Sounds like next step is to all play with new Amaya and see how far we 
can push this.
LK Concerned about stability of Amaya.
WC What version have you used? Latest versions are stable. I take minutes 
in Amaya, that shows how much confidence I have in it.
PJ Some of these things are more useful right away while others are long 
term. Today would be useful to merge output results and combine them and 
produce another report from that. If we come up with something so they talk 
the same language, that would be useful.
/* PJ discusses Bobby id's */
LK In terms of implementing this scheme, how likely is it you can or will 
do it?
BM We're taking things a step at a time. We've just begun working on 
interactive implementation. Saving history would go along with that. If we 
came up with a real spec then we would do it. In terms of "let's see how it 
works" prototype, it's less likely.
LK If we come up with a functional spec that says, "what info is stored and 
how do we point into the docs" then how to translate that into RDF or XML 
that should be straightforward.
BM Yes, those are the major problems.
HB I am concerned: each should be sufficiently generalized to apply to any 
XML application whose purpose is to deliver information to end users. Don't 
want to only to XHTML.
PJ If we can solve it for XHTML then we should be able to do it for any 
"ML" such as MathML.
LK In terms of robustness in the face of editing, if I only change white 
space, it would be good to have a notation where whitespace is not a 
factor. Don't want to CMN's hashing. Parses rather than deals with raw 
character strings.
BM Common to filter whitespace. Ignore case.
LK The order of attributes.
BM Now getting into a real parser.
LK Carry further, need closing tag in HTML?
/* PJ leaves. will continue participating on list. */
LK We're doing this on ER.
PJ I get that list.
/* WL arrives */
WL This a.m. proposed that rather that putting a logo at the bottom of the 
page, we put a statement in there re: this site.
WC In WCAG 1.0 Errata have a PICS statement one could use to make a 
conformance claim in metadata.
WL Pushing RDF-IG to actually use it. The proposal was that have RDF in 
XHTML that would validate. Right now it doesn't. It could contain name, 
conformance, and other semantics. I want to get started somewhere. If we 
can make a tool that says, "what's your e-mail address" it would put that 
in the document. To do it by hand is asking too much. If a tool to just put 
stuff in meta that would be something.
LK You want a meta-tag editor?
WL I have no clue how to do it. It should be done. Make it a technique in ATAG.
WC That's what Annotea is, a meta data editing tool, in some sense.
WL Note enough people use Amaya.
LK Sounds like you are building more accountability into this.
WL Yes. That's what people want. Govnt web site could decree use of that.
LK That's related to something I wanted to bring up on the list. Cuts 
across WCAG and ER. For accountability must be able to test if a page is 
accessible. I propose a WCAG checkpoint/guideline that you must test your site.
HB When independent people test a page rather than internal, if the page 
had "mailto:" then the test results could be sent back to them.
WL Whether a priority or convenience, we ought to do it. Anyone against the 
idea of creating a tool that will allow them to insert meta data into a 
document.
JR What if something more like a search tool and "access who" that would 
look for accessible pages.
WL Some people looking for, "is there anything for 4 year olds" others "i 
want to use an e-commerce site that won't give my screen reader a hassle."
LK If you come up with what it needs to do I'll write a Perl script.
WL Someone needs to find out why XHTML files can not validate with RDF.

Merging AERT and ATAG-TECHS
LK Any comments or questions on Jan's work?
JR Do people think it is the right way to do it? I don't want to do the 
whole thing if this is not what people had in mind. It's under checkpoint 
4.1, I have the repair there as well. Since it's relative there are 48 
subheadings. Each subheading has techniques within them. WCAG 1.1 has 9 
techniques.
LK You have authoring techniques under each checkpoint? Should we combine 
authoring and evaluation?
JR It's an arbitrary distinction. When author often looking at properties 
dialog, when evaluating may be a different dialog?
LK Save words to mix? Get rid of redundancy?
JR Might. Might complicate things. We have lots of info about dialogs and 
prompts.

$Date: 2000/11/07 23:01:51 $ Wendy Chisholm

--
wendy a chisholm
world wide web consortium
web accessibility initiative
madison, wi usa
tel: +1 608 663 6346
/--