2006-134 Meeting Minutes

PDF-UA 2006-134
October 17-18, 2006

PDF/UA Working Group

Face-to-Face Meeting & Teleconference

PDF/UA (Universal Accessibility) Working Group Draft Meeting Minutes
Tuesday, October 17, 2006,  9:00 A.M. - 5:00 P.M. EST
Wednesday, October 18, 2006,  8:30 A.M. - 2:00 P.M. EST
Conference Call Information
International Dial-in: +1-303-928-2600
U.S. Dial-in: 1-888-742-8686
Conference ID: 7083945

NOTE:  The meeting was conducted "live" on the Wiki.

See Logistics: Toronto, October 17–18, 2006 

October 17 

Roll Call

Present:

Duff Johnson, DJ - Document Solutions, Chairperson
Joe Clark, JC - Independent Accessibility Consultant
Ferass Elrayes, FE - Netcentric Technologies
Loretta Guarino-Reid, LGR - Adobe Systems
Karen McCall, KM - Karlen Communications
Greg Pisocky, GP - Adobe Systems
Neil Soiffer, NS - Design Science

On Conference Call:

Dick Herring, DH - Independent Accessibility Consultant

Approval of PDF/UA 2006-132 Agenda

Approval of 2006-132 Meeting Minutes

Report on / update and discuss [[PDF/UA - Action Items]]

Discuss options for v1.0 of the PDF/UA spec

Ferass Elreyes presented the options. 

2 options on how to structure.

 JC: Minimal set and optimal set. Minimal: all known disability groups. Something for everyone. Mandator y and optional - content related and disability related.

FE: Begin with something that would make any document that would be minimally accessible.

LGR: So what does that mean?

FE: It means things such as every image must have alternative text. If you have an audio you need to have it synnchronized. If you don't have headings, then it would be difficult to, but not impossible to navigate.

JC: It may end up that the minimal set is pretty strong.

KM: Samme approach as web design. Good navigation, alt text, annd headings. These are the basics to start.

DJ: When we say basic, we mean a document without a table for example.

JC: If we have a minimal set and you come across a document that has other elements, then you move to the next step up.

DJ: That's modules to me. These are content driven modules.

NS: DAISY took that approach, what it means to do math, what it means to do DRM, etc.

DJ: Seems like a way to deal with the PDF/A issue. Another goood thing about the modular approach, then it accomodates Neil's concerns. If it includes math, then the module is triggered for math.

NS: Same for multimedia.

DJ: Who would like to argue for priority levels at this point?

LGR: Let's start and then discuss what is mandatory and optional.

DH: Would mandatory and optional permit allowinng making some things conditional.

DJ: While it be structured in priority level terms?

FE: THe shalls, shoulds, and musts are implemented in terms of priorities in the W3C for example.

FE: If we want to releasee a version, we need to address all aspects of a document at least minimally.

LGR: That means, we have to have a module to address everything. You want to make sure if you go down the module path.

JC: We just need a module for known content types that have known accessibility issues. Minimal might be alt text, say "logo". Optional might be "actual text" in the example of this logo. If one were left out, then we could leave roomm to publish addenda on that.

LGR: We do have some unique problems. PDF for instance has issues regarding text.

JC: Then the minimum might only apply to single column text.

LGR: THe thing about text is what has to be there for different pieces of text. What are the rules for composing? 

 JC: That's reading order?

LGR: Sometimes reading order, sometimes grouping and associations.

FE: What does a minimal set mean if we are not going to priorities?

JC: Minnimal, general accessibility needs. All of the broad groups are in the set, and things that are unique to PDF, like reading order are in the minimal set. But add things like math, tables, etc. MMinimal set has sommething for everybody.

KM: Document structures with design layout, and documents without design layout. Tables for design layout.

JC: We could ban tables for layout.

GP: The shortcomings come from the fact PDF is a destination format not a source. People use tables for layout because they are the workaround for shortcomings in the authoring tools.

DJ: We have talkeed about priority levels which as a formal structure seem to be less preferable in this group. What I am hearing, instead of priority levels we adopt a certain subset for "simple documents" and add modules for more complex objects (like tables). 

JC: Simple document, then table module, forms module, etc. 

 DJ: What we have discussed and we seem to be reluctant to go the priority route per WCAG. But the ovverall model for this is what we are trying to pin down.

DJ: I feel as if we are arriving at some consensus here. Simple document model, additional modules that are content drriven.

JC: Let's not forget disability driven.

FE: It strikes me that the modules will still be coontent driven. 

 DJ: Does everyone feel comfortable with the broad strokes we have been discussing?

 FE: Do you see a module based standard being adopted by any government?
 LGR: By adding modules, and we should check with Betsy, we have created a sitution that people can create documents that meet version 1.0 but not conform to 1.1

 KM: As we evolve this, items may become obsolete.

NS: There is depracated stuff in PDF.

AI: Check with Betsy on this approach

JC: Joe suggested that time dependent criteria might be worth considering.

 JC: Proposed: The PDF/UA specification shall consist of modules defined by document content. To conform, a PDF document must meet the terms of all modules applicable to its content. The specification defines what constitutes general accessibility criteria and the modules address specific content types that extend beyond the general accessibility criteria.

Day 1: Lunch Time Break  

Neil Soiffer discussed and demonstrated MathML and MathML in PDF using Acrobat. Subscripts and Superscripts may not be readable, hence the Design Science popup view. Alt text only offers one take on an equation, MathML provides users with choice.

Presentation MathML (30 tags representing most mathematical notation), corresponds to what people type.

Content MathML tags that describe semantics, layout via conversion to presnettation via stylesheets, etc. 140 content tags with 12 attributes.

"math' is the rooot tag

semantics - ties together presentation content or other representations

maction associates actions with an expression

Summary of the tags, Token, General Layout Elemments, Scripts and Limit elements, Table Elements, Misc. Presentation Elements, Other = about 30 tags in all. 

NS: Proposal: All presentation elemments should be legal in PDF. MathMML attributes are stored as Atribute objects. Content elements must be inside "annotation-xml" element (inside a "semantics" element).

•     Doesn't always map well too the PDF in the document.
•     Can be cross linked to the presentation via id/xref attributes
•     Content can be used to povide a better audio description 

Maction is legal but does nothing.

LGR: You will need an atribute owner (perhaps MathML) then a type, number, string, etc. 

NS: Content annotation is a may.

LGR: Major motivation for going down this path, you need a finer mapping between the respresentation on the page and the PDF.

 NS: Proposed: The PDF that represnts a mathematical expression shall be tagged with a "formula" tag. Inside the formula tag, MathML's presentation elements shall be used tto make the math accessiblee The tagging shall be done in a manner that allows navigation of the substructure of the mathematical expression.

The Math MML may also contain content elements to convey sematic information in addition to notational inforation.

Top level tags: formula, math, semantics

LGR: We need to add to the table the lists of constraints.

NS: 30 tags and 50 attributes.

DJ: You have proposed adding a series of tags to the PDF Reference.

NS: The general issue is that there are some parts of MathML that  don't map well into the PDF structure.  You also can't get by with just content alone. So that has to be resolved.

NS:  Then there's the notion of simple math. Something like t is greater than zero. You probably would not bother tagging such an equation.

JC: Example 6.67 x 10²³ eV and 6.67 X 10^23 eV, neither is ambiguous. 

A demonnstration using two different screen readers the working group had access to indicated that the operator X ("times") was ignored by the assistive technology, strengthening the case for indicating even simple formulas / equations as math. 

NS: Proposed: All mathematical expressions shall be enclosed within a <Formula> tag. All mathematical expressions that can be represented  by MathML may be represented as MathML using a <Math> tag inside a  <formula> tag. All nonlinear mathematical expressions that can be represented by MathML shall be represented by MathML.

Soiffer/Clark

Agreed to Unanimously. 

For tomorrow, determine the names of the modules and come to conclusions as to what is in and out for each. Tables for example.

  Review Status of Draft Specification

2 conditions as indicated in the Valid Methods for tables.

1. The table structure is expressed in valid tag typs, attributes and valid table structure.
2. Each  

Day 2, October 18, 2006

Roll

Present

Duff Johnson
Joe Clark
Ferass Elrayes
Betsy Fanning
Loretta Guarino-Reid
Karen McCall
Greg Pisocky
Neil Soiffer

Phone

Mark Gavin
Dick Herring

Table discussion.

Need for Colspan debated. Joe Clark presented the following as an example of symantic benefits of doing so. http://joeclark.org/access/cinema/reviews/

FE Presented candidate algorithms for programmatic validation of table structures.

1. Associate data cell only with header cells at the top of its column and left of its row (including span attributes). All other associations must be expressed explicitly with PDF's id and headers attributes.
2. Associate data cells with all header cells at the top of its column and the left of its row. That is, accommodate all headers in rows/cols 1..m before the data starts.  All other associations must be expressed explicitly with PDF's id and headers attributes.
3. Associate data cells with all header cells at the top of its column and the left of its row. That is, accommodate all headers in rows/cols 1..m before the data starts.  All other associations must be expressed explicitly with PDF's id and headers attributes.
4. Define an optional "header level" attribute for TH cells that defaults to level 2 (1-based) for all but the first row/column.  The first row/column would default to level 1.  Scan row and column for closest TH at each level.  All of the headers at different levels are associated with the cell.
5. The row level algorithm
1. The list of row header cells is constructed according to the current HTML specification.
2. For each row that contains a header cell that was added to the list, if the row has a ROWLEVEL attribute of N, then the TR tags in the table are searched backward, looking for a TR tag with a ROWLEVEL less than N and greater than or equal to the stoplevel of the initial row.
3. If such a TR element is found, then the header cells in the row are inserted into the list. Note that a row may contain multiple header cells. Only the leftmost header cells (all header cells occurring to the left of any non-header cells) will be inserted. Then N is set to the ROWLEVEL of the row that was found, and the backward search resumes.
4. The search process is repeated until it is stopped by a TR element with ROWLEVEL less than or equal to the STOPLEVEL, or the beginning of the table, or a </THEAD> or </TFOOT>. If a row with a ROWLEVEL equal to the STOPLEVEL is discovered, that row's header cells are added to the list.

DJ: Let's review the alogorithms and select the best. Algorithm 1 is not sufficient.

FE: Headers and IDs remain a fall back position for these.

Algorithms 1 and 2 have been ruled out because they have been determined  by the working group to be too simplistic. 

FE: Demonstrated a complex table from a state highway administration.

NS: The ID solution permits one someone to mark everything in a table.

JC: An attribute Exclude= would prevent the chaining.

NS: Chain = False would be a simpler idea. Options are you go into a chaining mode or you restart the algorithm.

Proposal: New algorithm for tables (6). Joe Clark has also noted that chaining is in the PDF Reference already (p. 863) So the working group not adding to the specification. This definition is ambiguous because it does not indicate the order.

1. Add an attribute: "Stop Chaining" (True/False)
2. Look left and up to find the nearest row and  column TH.
3. If there is a Headers attribute on the encountered TH, then
1. add the header cells in the order in which they are listed (This algorithm will require more work - AI: LGR, NS, and FE will refine algorithm 6.

DH: Question concerning a summary attribute for tables analagous to HTML.

Upcoming Meeting Schedule

A face to face will occur during the AIIM Conference and Expo - April 17 - 19 -  Boston, MA.

November 1, 2006
November 15, 2006
December 13, 2006
January 10, 2007
February 7, 2007
February 21, 2007
March 7, 2007
March 28, 2007
April 11, 2007

Face to face to occur during AIIM Conference and Expo April 17 - 19 -  Boston, MA. Exact date to be determined.

Modules

PDF/UA shall be organized into a Framework of Modules.

The application of each Module is content-driven. All conforming documents will comply with the Framework and with all Modules corresponding to the document's content.

Framework

Define the relationship between PDF/UA and the PDF Reference 1.6

Statement of Conformance

Nomenclature for modules

Reading order

Artifacts

Language

• General Requirements
• Flickerin
• The use of color to convey information
• The use of contrast to convey information
• Best Practices / Informative Annex

If a document has content not covered by any Module, the Framework provides the minimum conformatnce requirements.

Minimum Conformance Requirements

• Logical Structure
• Reading Order

Modules

• Math
• Image / Graphics
• Artifacts
• Multimedia
• Forms
• Text
• Tables
• Annotations
• Navigation Elements
• Attachments

Acknowledgements

Thanks to Joe Clark for his hospitality and assistance serving as the Working Group's guide to Toronto and his the advance arrangements he made with the Gladstone Hotel.

Thanks to Adobe Systems Canada and the Toronto office for hosting the meeting and providing meeting facilities.

October 18