Universal Design and the Grid

[working draft]

Alfred S. Gilman
with input from:
Gregg Vanderheiden
Gottfried Zimmermann
Trace R&D Center

College of Engineering, University of Wisconsin-Madison
2107 Engineering Centers Bldg., 1550 Engineering Dr., Madison, WI 53706
http://trace.wisc.edu | info@trace.wisc.edu

Table of Contents


The Trace Center has an EOT-PACI activity in the area of Universal Design for Advanced Computational Infrastructure. This document collects information based on our experiences in Universal Design that may be of value to the Grid Computing Environments Working Group or more widely in the Grid Forum.

Universal Design posits that products and services should be designed for the widest possible diversity of users, particularly as it affects demands on user capability. Likewise, that the remedy applied to accommodate special needs should be, and can often be if you look for it, a design feature yielding benefits across a far wider user base. A functional limitation is not a disability until there is something you can't do because of the functional limitation. Universal design aims to prevent functional limitations from turning into dis-abilities. And it aims to capitalize on the added depth of understanding of HCI issues that can come from looking at the design from a wider range of viewpoints.

The Grid is an attempt to expand the possibilities for human benefit from distributed and remote computation by dint of architecture of common practices. The possibilities are great, given the radical advances in both computer and network performance that are coming into use. This potential could be an engine for making information services more broadly available and accessible to a wider range of people, in particular people with disabilities. But this outcome requires attention to certain design issues early in the process and from the ground up in the architecture.

This white paper collects lessons learned from the area of Disability Access to Information Technology and Telecommunications. For example, one of the lessons learned is that the requirements of access by people with disability and access from small mobile devices are very similar. Thus paying attention to scalability issues in the characterization of services and resources will help people with screen readers, but not just people with screen readers. It will also help the Grid reach and be used by people via a wide variety of internet-interoperable devices, including mobile devices. The paper also outlines, in a provisional fashion for discussion, the relevance of these lessons learned to elements of the Grid architecture, starting at the Human/Computer Interface found in Grid User Environments.

Status of this Document

This document is a rough draft for a possible Grid Forum FYI which the Grid User Environments would recommend for Grid Forum publication after review and editing in the Working Group.

However, this is but one of many ways that the material here, and other knowledge from the Universal Design community, could be incorporated into the work of the Grid Forum in many different ways. The Working Group should feel free to consider unbundling this information and incorporating portions of this material as sections in documents with various themes and scopes however the Working Group decides to organize its work.

Executive Summary

Three points can focus our view as we organize Grid User Environments work in the Grid Forum, and wonder how to apply Universal Design in the process:


The Grid can grow or die by its usability. The usability innovations of the World Wide Web made Internet use accessible to many more people than before; and interest in the Internet exploded. A Grid computation is a potentially a complex thing, and generates voluminous output. The user will need to be able to book, monitor and control the computations being done on their behalf without any background in the house culture of the performing organization(s). And use the results efficiently in their own work environment. This will require new progress in the area of common practices. The discussions of opportunities in the User Environments and User Services Working Groups, below, suggest some of the ways how the accessibility and usability of Grid Computation can be maximized.

Early Adopters

Over the years many technologies were initially created to solve problems for people with disabilities (PWDs), such as the typewriter and the long-playing record. Blind people have been using personal digital assistants for years; deaf people are keen on video chat. There are lessons learned from their experience. Those who will not learn from history will be forced to repeat it.

linear version of table 1

Access Problem

What's Wrong

Current Fix

Better Fix?

Grid Impact

missing ALT text

broken page logic

complain to webmaster

SVG leaves words as text

device-independent resources

"click here" links

implicit context dependency

change visible link

mark up dependency

clear experiment controls

top links tank trap

true page structure not coded

skip-navigation hyperlink

criticality markup

auto-summarize trouble to pager

conversation between ASL and voice speakers

no common language

on site personal assistance

remote personal assistance

broadband networks, scheduling

hearing impaired and video

flow depends on dialog

closed captions hardwired

SMIL gives bundling capability

monitoring simultaneous simulations

blind users and GUI Aps

logic implicit in layout

guess from labels

haptic mouse

interaction with multi-dimensional information

Web Reform

Web users with disabilities and developers of novel web applications all find shortcomings in the current generation of Web protocols and formats, after the era of explosive growth we have just been through. The Web is in the process of rolling out a second generation of technology based on the streamlined foundation of XML. This includes scalable graphics formats so Grid results can be printed in wall scale and thumbnail scale. It includes multimedia integration capabilities for observing distributed computations as they compute. The Grid will for market reasons want to be highly usable when reaching customers over commodity last mile technology. Sharing in the ongoing reform of Web technology will reduce the amount of Grid-peculiar technology which has to be created and supported. The goal of course is that Grid innovations will be taken up and supported by the commercial marketplace. In a domain such as this where there is so much commercial innovation, if one is not playing the game that they are playing it is unlikely that the results will be competitive and take off. Current Web reform efforts of particular interest for Universal Design of the Grid User Environment are noted in a section on Current Technology Review. After some Analysis we then turn to how these technology trends create Opportunities for Grid Working Groups.


This paper is about Universal Design and the Grid. First, let's recap a little by way of a summary of each.

Universal Design Applied to Information Technology

Step back, look around

Whenever someone discovers a new problem, it is time to look around and see who else may be experiencing similar problems. And by the same token, every time you figure out a way to solve a problem, it is again time to check if you have taken in the full extent of the opportunity before you. Quite likely there is a device or technique very much like what you have worked out that would solve related problems for many more people. The following table traces operational requirements that are make-or-break issues for people with disabilities to other situations where the capability is also needed.

linear version of table 2


Disability Related Need

Situation Related Need

Operable Without Vision

people who are blind

people whose eyes are busy ,e.g., driving a car or phone browsing, or who are in darkness.

Operable With Low Vision

people with visual impairment

people using a small display or in a high glare, dimly lit environment

Operable With No Hearing

people who are deaf

people in very loud environments or whose ears are busy or are in forced silence , e.g. in a library or meeting.

Operable With Limited Hearing

people who are hard of hearing

people in noisy environments

Operable With Limited Manual Dexterity

people with a physical disability

people in a space suit or chemical suit or who are in a bouncing vehicle

Operable with Limited Position or Reach

people who use a wheelchair or have limited reach

people who are out of position or have multiple devices to operate

Operable With Limited Cognition

people with a cognitive disability

people doing a task for the first time, or having to perform multiple tasks at once.

Operable Without Reading

people with a cognitive, language or learning disability

people who just have not learned to read this language, people who are visitors, people who left reading glasses behind.

Proven Strategies

Over the years, some patterns have emerged as to general strategies which have proven effective in delivering broadly usable services and products. The following table recapitulates some of the high spots. For a more detailed review, see the On-Line Universal Design/Evaluation Tool at the Trace website.

linear version of table 3

Basic Access Principle


How –General

Make all information perceivable, including keys & controls -status & labels

  • Without vision
  • Without hearing
  • Without reading, with low vision or cognition
  • Without color perception
  • Without causing seizure

Information which is presented in a form that is only perceivable with a single sense e.g., only vision or only hearing is not accessible to people without that sense. [Also not accessible by anyone using a mobile technology that does not present all modalities such as a phone – or automobile audio-only browser.]

Make all information available either in:

a) presentation independent form e.g., electronic text which can be presented in any sensory form,


b) sensory parallel form where forms suitable for presentation in different sensory modalities are all provided in synchronized form e.g., a captioned and described movie,


Provide a mechanism for presenting information in visual, enlarged visual, auditory, enhanced auditory louder and if possible better signal to noise ratio and, where possible, tactile form.


- To meet these requirements, text formatting must be electronically readable and presentable without vision.

Provide as least one mode for all product features that is operable:

  • Without pointing
  • Without vision
  • Without requirement to respond quickly
  • Without fine motor movement
  • Without simultaneous action
  • Without speech
  • Without requiring presence or use of biological parts touch, fingerprint, iris, etc.

Interfaces which are technology or technique specific cannot be operated by individuals who cannot use that technique, e.g., a person who is blind cannot point to a point in an image map; some people cannot use pointers accurately. [Also not accessible to mobile users who are using voice to navigate for example]

a) Make all interfaces controllable via ASCII/UNICODE input and output, AND

b) Have all text output voiced or compatible with device that will voice it, AND

c) Make all input and displayed information non-timed, or allow user to freeze timer or set it to very long time, 5 –10 seconds to do single action, 2-4 seconds to stop action, AND

d) Have at least one mode for achieving each and every function of the product that meets the following conditions:

  • No simultaneous activations
  • No twisting motions
  • No fine motor control required
  • No biological contact required
  • No user speech required; AND

e) If biological techniques are used for security, have at least two alternatives and preferably a non-biological alternative unless required.

Facilitate Navigation

  • Without sight
  • Without pointing ability
  • Without find motor control
  • Without prior understanding of the content
  • Without the ability to hear

Many individuals will not be able to use alternate access techniques if their layout is too difficult to understand.

Many individuals will not be able to operate products, such as workstations, with sufficient efficiency to be competitive if navigation is not easy.

a) Make overall organization understandable e.g. provide overview, table of contents, site maps, etc.

b) Do not mislead/confuse. Be consistent in use of icons or metaphors. Do not ignore or misuse conventions.)

c) Allow users to jump over blocks of undesired information, e.g., repetitive info , especially if browsing auditorially.

Facilitate Understanding of Content

  • Without skill in language used on product
  • Without good memory
  • Without background

People with cognitive difficulties may not be able to access and use complex devices or products with language.

Many others may find that they are unable to master alternate access techniques if layered on top of complex interfaces or content.

a) Use the simplest, easiest to understand language as possible, appropriate to the material.

b) Say things both through literal statement and suggestive imagery; spell out your message and also act it out.

c) If unexpected languages are used – be sure the language used is identified to allow translation.

Compatible with Assistive Technologies commonly used by people

  • With low vision
  • Without vision
  • Who are hard of hearing
  • Who are deaf
  • Without physical reach and manipulation
  • Who have cognitive or language disabilities

In many cases, a person coming to a task includes the assistive technologies they have with them. If they cannot use products directly, it is important that the products be designed to allow them to use the tools they carry with them to access and use the products. This also applies to mobile users, people with glasses, gloves or other extensions to themselves.

a) Support standard points for connection of

  • audio amplification devices
  • alternate input and output devices or software)

b) Do not interfere with use of assistive technologies

  • hearing aids
  • system based technologies

c) Support use of wheelchairs, walker, etc.

Users with Disabilities Check Quality

The following paragraphs summarize quality factors that are valuable to the Grid, and where disability access scenarios -- walking through them in design and exercising them in evaluation -- help to make sure that the Grid is actually getting there.

Eyes-free and meticulous publication

Ensuring usability eyes-free is a quality check on how well published the Grid-served resources are. Have you really put everything you know at the disposal of other users? If you haven't supported your resources with schemas or other forms of documentation, especially machine-interpretable forms, you may not be adding the value to the Grid that you think, or that you could.

Hands-free and expeditious control

Ensuring usability by the motor impaired is a quality check on how well structured decisions trees within the user control of services are. User with few actual command symbols available to them walk longer protocols or decisions trees to get the system told what to do. This stresses the logic of the interface design, and will turn up weaknesses in this area.

Ears-free and process clutter

Ensuring usability by the hearing impaired is a quality check on how well structured the real-time processes are, which becomes critical in exception handling architectures for a distributed real-time experiment being run by a collaborative group.

Cognitive checks cure pedantry and accelerate automation

Taking steps to manage the cognitive demands place on the human segment of the system can, in the first instance, make more science and computation accessible to a broader public. "Dumbing it down" is not dumb, when practiced as a check on gratuitous obscurity.

Minimizing cognitive burdens is also critical on the path to where the computational segment can do more and more for the same expenditure of labor in the human segment. In aerospace applications, the human is often the limiting element is the capability of the system, and man-machine systems are rigorously scrubbed for MANPRINT factors. This is not quite the same case, here. But the cognitive demands of using the Grid _are_ likely to limit the _capacity_ of the marketplace to absorb Grid services. The more people can understand how to do it, and the more people can do it with less mental sweat, the more computation the Grid can sell.

Technology Watch: Promise and Problems in IT practice and trends

A few quick examples

Here we walk through the examples presented above in the Executive Summary, tracing from problem detection in IT use by people with disabilities through to Grid gains if a better fix than the current fix can be made part of the endemic base of Information Technology.

linear version of table 4

Access Problem

What's Wrong

Current Fix

Better Fix?

Grid Impact

missing ALT text

broken page logic

complain to webmaster

SVG leaves words as text

device-independent resources

"click here" links

implicit context dependency

change visible link

mark up dependency

clear experiment controls

top links tank trap

true page structure not coded

skip-navigation hyperlink

criticality markup

auto-summarize trouble to pager

conversation between ASL and voice speakers

no common language

on site personal assistance

remote personal assistance

broadband networks, scheduling

hearing impaired and video

flow depends on dialog

closed captions hardwired

SMIL gives bundling capability

monitoring simultaneous simulations

blind users and GUI Aps

logic implicit in layout

guess from labels

haptic mouse

interaction with multi-dimensional information

Missing ALT text

HTML provides the ALT attribute on IMG elements so that if images are used to control the exact appearance of buttons, logos, etc, there is still standard text available so text-to-speech technology can be used as in screen readers that blind and visually impaired computer users employ. Otherwise the words embedded in the pixels of the GIF or JPEG image are not recoverable as text to be spoken by the screen reader. This leaves gaping holes in the logic of web pages, as when a hyperlink contains nothing but one button image. If there is no verbal content, the blind user is left without a clue as to where the link goes.

Populating the text into these attributes, however, is one of those "documentation" tasks that often gets honored in the breach. Many websites leave a lot to be desired in the completeness of this annotation. Presently there is nothing to do but complain to the webmaster. The emerging Scalable Vector Graphics technology, however, lets the designer control the appearance of a logo or button precisely, have it look good on large and small displays (which present raster images don't) and the text is still there as text within the SVG so that it is searchable and readable by screen readers. The Grid will be very interested to have some sort of better scalable graphics technology than what is available today, because the Grid user needs displays from wall size to wallet size all to have access to their work.

"Click here" links

Many authors, to make the hypertext comprehensible to novices, don't trust to underlining or other stylistic effects to tell the user where the sensitive text is, but rather spell it out plainly by using the words "click here" as the sensitive text of a link. This minimizes the chance that a user won't know what to do. On the other hand, as a blind user tabs through the links of the page (see "why blind users tab through web pages, below) what is read is just the highlighted text in the focused element of the user interface. Hearing "click here" gives no clue at all as to what happens if you do. Here we are back with Alice in Wonderland with the bottle labeled "Drink me." The identification of what clicking there is clear enough in the full context of the running text around the link element, but the link element markup doesn't tell you what or where to read to figure it out. The current fix is, if you can find them all and persuade them all, to convince authors that the sensitive text should be chosen as a fit label for the action that results from clicking on the hot link. This works for those authors we can reeducate. A possible better fix is to let the "click here" be the actual active text range but to have the hypertext format itself contain information as to what text in the environment of the active text is explanatory of the link action. For example, extend the LABEL element so that it applies to hyperlinks as well as form controls. There are various ways this meta-information could be encoded in the medium. For table cells, where this problem is endemic, various attributes were added in HTML 4.0 that capture the description:instance relationships in the table. In any case, the Web designer community is working around the lack of a way to say this in the HTML with scripted annotations that respond when the mouse is over an element. But burying this information in a script hides it from the assistive technology. A more systematic, declarative approach would be more successful universal design. The HTML Working Group in the W3C is opening the floor for functional reforms with the planned development of XHTML 2.0. The right kind of this sort of informative linkage in HTML will benefit Grid user interface designers through robust mechanisms to ensure that controls are understood.

Top links tank trap

This issue is discussed at greater length in a later section. As it is, many web pages have many links lined up along the top margin which are quickly skipped by visual users who jump to the meat of the page. The eyes-free user, however, does not have the broad overview of the page content to rapidly locate the meat of the page and also start reading there. They tab through the links in hypertext sequence. Because of the repetitive nature of the links in a header, it can be hard to know which page of a site you are on. Presently the fix is to insert a dedicated hyperlink from early in the tabbing sequence to a reasonable point to start reading the meat of the page. Future revisions of web content formats could have each context able to tell you what in its content is more and less critical, that is, less or more expendable. This would also be extremely valuable for automatic summarization which finds value in professional communication and reaching mobile devices.

Conversation between ASL and voice speakers

Presently sign-language interpretation for conversations between people who use American Sign Language or other sign languages and vocal speakers who haven't learned ASL is labor-intensive. Interpreters work in teams because of the pressure of the task. Because the demand is thinly spread, resources are scarce and have to be scheduled far in advance.

Serving interpretation needs remotely from a central site would allow more efficient utilization of the interpreters and greater availability (shorter scheduling lead times) for those needing service. The Grid has one of the basic requirements for remote service in broadband networks. The scheduling and basic ordering agreement arrangements would have to be worked out, however. As voice and gesture recognition technology improves, the Grid can be the way such capability is made available as well. This service function is the topic of an ongoing line of research within the UD/DA for ACI activity at Trace. The Grid would benefit from the quickturn scheduling infrastructure for debugging and crisis management applications, and the translation services model could be extended to covering international collaborations.

Hearing impaired and Video

Most video programs are highly dependent on their sound track for comprehensibility. People with hearing impairments need some relief. Closed captioning is the best current practice. However the media representation of the closed captioning is tightly bound into the video medium. This does not support third-party captioning well, or more general patterns of information flow such as would arise in multi-point collaborations among Grid users. The Synchronized Multimedia Integration Language is entering its second generation. This allows for flexible bundling of streams of information. It could prove advantageous for the creation of experiment consoles for distributed Grid computations where there is video being sourced from multiple points and the user needs a flexibly integrated presentation of the ensemble.

Blind users and GUI Aps

Screen Readers have caught up now so that applications designed on the assumption of the dominant Graphical User Interface windowing system are about as usable (by screen reader) as their DOS ancestors were five years ago. This is good news for job integration of people with visual disabilities, but it is not that good news. The effectiveness of this situation is highly dependent on the application developers going through and assiduously labeling elements with textual labels for eyes-free comprehension. But sometimes the logic of the interface layout is truly two-dimensional, and labels do not capture it all. There is anecdotal evidence from blind workers who have been able to work with research examples of haptic (force-feedback) mice that the block layout of the visual interface is easy to track and understand with this additional information. The experiments of computational science, which is an important user community for the Grid, involve many variables at once. Any unused senses are a missed opportunity in creating an interaction environment for research computational scientists. There is a common interest in proof-testing this technology, and the scientists stand to benefit ultimately if it can be raised to the level where visually impaired individuals put it to widespread use.

Progress and Problems on the "Top links tank trap."

Why blind users tab through Web pages

In browsing with a visual display, we skim. We glance across the page and pick something to read. We don't read it all. Without some skimming strategy, auditory browsing is left to just read the whole thing. This can be a time-consuming process.

One strategy that is very commonly employed to alleviate this situation is to leave the screen reader in a mode where it reads what is highlighted on the screen, and then use the TAB command to cycle through the hotlinks and form fields of the page. This capitalizes on two factors that happen to be present in the visual Web environment. First, these active elements are of special interest. The basic orientation questions are "Where am I? What's there?" and "What can I do?" There is always a special interest in opportunities to do something. Hot links offer the option to do something. The second factor was that this special interest was supported in the existing software interfaces by TAB navigation. Borrowing a pattern from forms-oriented user interfaces, the TAB key was used to iterate between opportunities for the user to do something. In the original application this was form controls. The principle was extended to hyperlinks.

Enter terminal tedium

As the Web became immensely popular, the masthead of the typical Web page grew and grew. Web sites became matters of intense ego competition among candidate destinations, so it was important to have the sponsor's logo on all pages to develop brand awareness. Personal computers gained screen resolution so that more and more logos could be arrayed across the top margin. Sites developed more powerful utility functions, so the battery of tabs or services shared across the headers of many pages grew.

In the audio-browsing, tab-the-links strategy, this evolution of the page header raised a formidable barrier across the entry to many pages. As the number of links involved in pro-forma preliminaries grew and grew, getting to the meat of the matter go to be a more and more tedious process. The consumers were left asking "Where's the Beef?" The eye could skip rapidly to the headlines and high spots in the center of the page, but the screen reader only had the markup language view of the page structure, and was left plodding through links in the order of their appearance in the HTML Text.

A simple link does the trick

When Tim Berners-Lee grabbed SGML markup technology to create a simple hyper-media format, and bound it to the concept of a universal class of link targets or jump destinations, he unified intra-document and inter-document references in the process. References could be to another document or to another point in the same document. In the process of making the system regular this automatically included references to a defined point in another document.

The American Council of the Blind (ACB) was an early adopter of a simple technique enabling a site to provide both rich identification via hyperlinks in the masthead and expeditious motion to the meat of the page. This they did by inserting a hyperlink early in the textual order (tabbing order) of the page which skipped the preliminaries and got straight to the meat of the matter. This technique is identified in the W3C Recommendation on Web Content Accessibility Guidelines. Some such capability is also emphasized as needed in the draft U.S. Federal standards for the accessibility of Federal websites.

Digital Talking Books provide a friendly tree walk.

The Graphical User Interface (GUI) stabilizes the user's grasp of context by wrapping the immediate focus of interaction with a lot of data from the session context. Verbs are presented in menu trees, sessions are represented by window sets with portions of all windows visible if cascaded presentation is selected.

Audio display simply lacks this shared buffer of context cues. All one has in an eyes-free interaction environment such as is presented to screen-reader users is a stream of language buzzing by the ears. The context is all in what is held in the user's head. To make it easy to remember how to recover other views of the session from the context of what is happening right now, it makes a big difference how regular the structure of the context is. If the contextual setting can be framed against a backdrop of easily remembered patterns, it is much easier to recall and apply.

The table of contents tree metaphor, such as employed in books, is one such friendly walkable structure. In fact, in converting books for audio use, library services for the blind clean up the table of contents structure to make sure that it provides an informative and navigable topical index to the content of the book. Special features in the markup of digital talking books, and special instructions to the producers of such books as to how to ensure a quality table of navigation are described in the documents of the DAISY Consortium. The attention given to these features in this project demonstrates how important this can be in an audio-only interface.

So why can't we do better?

Looking at the graphic design of many web pages, they certainly have a "departmental" structure: recognizable parts with distinguishable (and not that hard to remember) functional roles. The masthead serves to remind one where the service is coming from. A common feature is an outline of related destinations reachable by hyperlinks, often placed running down the left margin. The right margin tends to attract use for asides: sidebars and promotions of other points of special interest. The foot of the page picks up administrivia and other matters that are of interest to a minority of readers but critical for those who are interested.

Today's HTML doesn't get to that level of understanding the page structure.

Masthead navigation bars wind up being coded as a sequence of elements without an enclosing element defining the header scope. Even where headers are used in a disciplined hierarchy, H2 elements appearing "within" the scope introduced by an H1, etc., there is no markup delimiting the scope of applicability of the headers. They are just markers salted into the flow of the hypertext, without associated containers the way there are, for example, in LaTeX.

The left-margin navigation bar is often implemented within a column defined in a table used to provide a layout grid for arranging components on the page. Because a column is a virtual structure in an HTML table, and not composed of elements contiguous in the hypertext, it does not have the sort of recognition value to a web browser that is needed to support identification as a functional unit.

One could guess. That's costly and risky.

There are many heuristics that could be applied which would increase the likelihood that an abstraction of the parse tree which the browser were to present to the user as a table of navigation for the page would be helpful. But in order to get a structure that was not laughable in most cases would take attention to multiple indicators, and even then there would likely be frequent cases where the constructed outline was useless, or at least laughable.

What's a User Agent to do?

The User Agent Guidelines Working Group in the W3C/WAI is struggling with this issue as we speak. Given the state of Web authoring practices, there's no easy algorithm to add in the browsers to achieve top-down orientation and navigation. With the current state of browser practices, there's no simple marking that the authors can adopt to gain the full structural effect. [But they should continue to provide a skip-navigation link at the top of the page to bypass the pro-forma preliminaries and get to the meat of the matter.]

One possible strategy

One way that this could work is if every object (XML element instance, subtree in parse tree) is able to inform you as to its dependencies; and every context is able to inform you as to its priorities. Dependencies: these are places where elements of the context are critical for the correct interpretation of the object, as headers may give critical meta-information about a table cell, labels for form controls, etc.

The priorities could take the form of criticality rating on the contents of the object. For example, the story that gets the lead headline in a newspaper can be afforded the top criticality all by itself in that page design. But runner-up position probably includes two other stories. The masthead and copyright mark are pretty well down toward the tail of this distribution. This is not necessarily a linear list, a total ordering. It could rather be a graded rating of what details could be pruned first, up to what critical items or connections needed to be preserved in summarization at all cost. Part of the motivation for this is that if the author has conceived the content as a Web page, they are perhaps more likely to understand the question if you put it to them which objects in their design are the most critical to save if it must be reduced, rather than asking them for a hierarchical container tree that reflects their design thinking. Often for Web pages the design concept is graphical. The arrangement of the objects and the interactions that this creates can be as important as just a list of the contents. The clusters of content objects that interact can be overlapping. There is not necessarily any way to allocate an object uniquely to a parent sub-layout that conveys the compositional structure of the design. Books, which are carefully composed to be comprehensible through a linear reading from front to back, can have their compositional structure mapped into a tree without missing a lot. Graphic layouts are richer graphs than that, and don't necessarily admit a clear explanation within the limited medium of a Table of Contents tree.

Visions of what Could Be

Let us now raise our sights a little, and walk through three example scenarios of possible future advances in disability access to information technology.

Assistance on Demand

The surprise visitor

Suppose you are expecting a small group of visiting scientists in your area and you want to talk to them about the Grid-hosted experiments that you have been conducting. The party arrives, and unbeknownst to you, one of them is Deaf or profoundly Hard of Hearing. It's too late to obtain the services of sign language interpreters; they are all booked two weeks out in advance.

The open, extensible conversation

As the Trace Center demonstrated at SC99, with the bandwidth of Internet 2 it is possible to pipe in sign language interpreting service from anywhere on the high speed network. This allows service to be provided from central locations, improving queuing performance to up both utilization of resources and quickturn access to service.

Over time, technology will improve so that more of the translation burden can be borne by computers. Real-time speech to text is in widespread use a) for limited vocabularies in telephone service contexts and b) where the recognizing software is trained on a particular speaker's vocal characteristics and only recognizes the speech of that individual. Services delivered by man-machine systems are extending the envelope of services available today.

Ruggedizing the Grid collaboration capabilities until there is always room to patch in more services on short notice will make the recovery from this surprise an easy matter.

Problem resolution demands unpredictable things

Of course, accommodating this special need is by no means the only scenario arguing that the Grid should tool up for quick turn-around construction of user-level services composed from Grid capabilities. Crisis management is one of the areas where massive computations may be desired, and there is little or no warning that the service is required.

Integration Problems require co-accessing experts

The Grid is about integrating far-flung and potentially dissimilar resources to create a composite capability. Sometimes there will be problems getting the parts to work together. When this happens, it often takes expert knowledge of the parts on each side of a problem transaction in order to diagnose and resolve the problem. And the experts probably have to interact, perhaps even in real time, to resolve the matter. Here the critical factor is often getting someone who knows each side together; not so much exactly who you get to represent either side. This is where an on-call pool of potential consultants could be useful. Some of the resources on The Grid will be novel codes and lightly used. But the Grid lets them be used from anywhere. So lightweight means to achieve some sort of support are of interest.

What the Grid could do


Define classes of booking problems over a wide range of response time requirements. Architect the resource capture protocols for all these levels of urgency.

User Services

Manage the standing-order procurement of remote interpretation service.

Create a Grid-Wide information base on consultant-capable personnel resources.

Information Services

The hearing problem is on record. It is not make public knowledge but the dialog between visit control of the two organizations carries this information along so that although the host scientist had been unaware of his visitor's hearing condition, the host administrative staff is alerted and makes the necessary arrangements in advance of the visit.

High-Capability Compute Factories

Offer machine cross-modal translation services beyond what the commodity technology platforms of the day can offer.

The portable Point of Sale

A deaf-blind customer pushes her grocery cart into the checkout lane, gropes the conveyor belt for a moment to make sure where it is, and unloads the cart onto the belt. The customer's Braille note-taker and the Point of Sale terminal have already discovered one another by BlueTooth short-range networking. As the checkout clerk rings up the items, they are posted into a living-document electronic register tape which by data synchronization is kept current in the Braille note-taker as well. The customer reviews the itemized bill of sale at their own speed and can move back and forth through the portion that has already been rung up.

Once the last item is rung up, the terminal messages its note-taker correspondent to double-check for a savings card number. The consumer picks the merchant out of a list of such accounts she holds, and the Braille Note-taker messages back to the store system with the details. This causes an adjustment to the bill.

At this point the consumer selects an account from which to pay for the transaction and approves the transaction. The Note-taker inserts the account identification information for method of payment, applies a digital signature that the financial institution can authenticate, and the deal is done. Faster than with a seeing, hearing customer that you have to do all this by voice and gesturing at the buttons where the user can't find "yes" on the card-swiping device.

An extensible Grid

What's happening here? The store automation understands Grid computing, so that it can JustInTime accept a new node into its computational grid and send work to the new grid node for processing.

What has been done is that the task of conducting the point-of-sale dialog with the consumer has been migrated at the last minute from the installed point-of-sale system to the computational node that the consumer brought with them. This means not only that they have to have the ability to establish data communication on a casual, walk-up basis; but that the transaction has to be defined in a portable fashion. Likewise, the mechanisms have to be in place for the rapid establishment of a trust network binding the transaction processing by note-taker, store computer, and bank computer. The consumer has to be able to trust that their device receives a full and fair representation of the deal, with the same prices offered to others on the visual displays, and the store automation has to have a basis to trust that the data communication it receives from the walk-up device represents a legal commitment by the consumer that the bank will honor.

What this requires is that the store network be set up to be open to accepting the customer-supplied equipment as the actual terminal equipment to conduct the point-of-sale confirmation dialog. In addition to the installed hardware devices, the system needs to be configured to recognize and deal with any accessor capable of interacting under the terms of some sort of an Alternative Interface Access Protocol. The data communication part of this scenario is clearly coming fast. The portable transactions -- at the bill of sale level of complexity -- are only as far away as XML Forms.

This scenario is only a vision of how it would be great if things worked. On the other hand, there is already a standards committee forming up to try to define the protocol that would enable this flexibility for deaf-blind, high spinal-cord injury, and other people with peculiar computer interfaces to use their adaptive interface to a personal digital assistant and have the digital assistant talk computer-to-computer with a wide range of IT devices used in service to the public.

The trust extension platform is not so clear where it is coming from. The Grid has a similar problem in setting up trust tents over service delivery maps that cross administrative domains.

What the Grid could do

The key here is probably the work in the Security Working Group on how to have networks that are both trustworthy and open. If The Grid shows that it works, the commercial world will have a model to follow.

Haptified Imagery

Haptics is a term for that which you can grasp, literally; sense with the hand. In particular it has come to be the term of choice for information supplied by sensing the force required to move things around, such as a stylus or mouse.

Computational science is interested in anything that can be done to add dimensions to the virtual world that the user interacts with. Science problems generally are multi-dimensional. Their dimensionality exceeds that of the virtual worlds that we construct with even immersive technology. A slice of the problem space has to be selected for direct representation in the interaction medium, and something important inevitably gets left out.

People with visual disabilities have reasonably good access to computer text that represents a narrative flow. Other information that contains an intrinsically spatial element, such as spreadsheets or map data, do not map so well into an interface that is entirely dependent on audio or a few Braille cells for output. It is possible to get tactile graphics printed out with special printers, but this is slow and has low resolution.

There is evidence from blind programmers who have worked with research devices that haptifying the organizing infrastructure of the GUI medium is highly effective. Little speed bumps as one enters and leaves dialog panels and transitions between menu choices give just the added level of orientation to the planar structure of the GUI desktop that it takes to keep one's mental place in the organization.

Haptic mice have broken the $100 barrier on the retail market. But the mapping to haptic effects is not there in most of the computer interfaces and documents that are available today. The affordability of these devices make haptics a hot topic for research into access to scientific and other graphic information. The Science Access Project at Oregon State University is working on integrating haptic effects with graphic presentations encoded in the Scalable Vector Graphics (SVG) format.

Promising Evaluation Resource

And example of a data resource that would be a good vehicle to investigate and demonstrate the potential of haptic imagery is the Carnegie Collection of human embryos. This is an archival dataset similar to the visible male and visible female, but it covers the early embryologic development of a human. The basic medium is volume data from imagery, but the collection is rich with annotations because it has been intensively studied and annotated by experts. Thus even if the resolution of the image information were greatly reduced, the binding of the annotations to the gross morphology of the embryos would provide a valuable resource for the education of blind students.

Human factors experiments with this collection would add to what is known about how dense haptic information can be in the plane and still be resolved by the user, and not obscure the low-frequency information by dithering caused by the detail. Outlining techniques and pseudo-relief can be compared. This would lead to advances in the state of the art in transcoding general image and volume data into haptic-appropriate form.

Once some progress had been made in the haptification of information using blind subjects and the embryo archive, it is possible that the haptics could be ported back into the medical application with good results. Where a surgeon has to review the results of fusing data from multiple sensing modes, there is a lot of information and the screen variables can get overloaded. For example, haptic cues could be used to indicate the presence of annotations, so that the pure imagery from the patient did not have to be obscured with highlighting or reference marks.

What the Grid could do

The Grid can help push this thread of technological progress along in a variety of ways. By integrating Big Data archives such as the Carnegie Collection, the grid makes available world-class resources not only to medical education but to interface research. In addition, the haptification process is at present an exotic art, not widely understood. The service linking capabilities of The Grid will make it easier for services such as the haptification of interfaces to be economic as a specialist industry.


HCI Fundamentals and PWD Failure Modes

Here we step back and approach the failure modes that people with disabilities sometimes encounter in attempts to use inaccessible human-computer interfaces.

The grok-locked loop of Web wandering [D]

The cycle of human-computer interaction is a rapid alternation between half cycles that are performed by the computer alternating with half cycles performed by the person. The computer accepts input through its command device(s), computes according to the instructions of a stored program, and generates and adjusts what it presents to the user via display devices as a result. The user senses and perceives what is presented via the display devices, cognitively recognizes things like language and other iconography within what is perceived, and uses this to adjust what they recall as what is happening in the dialog They update a state variable we could call "Que pasa?". Then, based on this concept of the flow of the interaction, the person projects an expectation for what is likely to happen if they perform one or another of the available action options. Comparing the expectations for different action options, they decide on what to do; and act, applying input to the computer's command device(s). The cycle repeats.

The term grok-locked loop alludes to the concept of a phase-locked loop used in demodulating modulated-sinusoid signals. In a phase-locked-loop demodulator, the modulation is observed in a floating reference frame of amplitude-perturbation and phase-perturbation coordinates. In order to observe in this coordinate frame, a local reference oscillator must be created which is phase-aligned with the carrier of the arriving modulated signal. Demodulation quality is greatly enhanced by this technique, but it depends critically on the ability to maintain phase lock on the carrier in the arriving signal. This lock is maintained by a control loop whose objective is keeping the perceived phase-pertubation centered around zero. The user's situation awareness or concept of process context in human:computer interaction is the analogous local reference frame in which system behavior is observed, like the phase-locked perturbation coordinates in the phase-locked loop receiver.

To grok is to "grasp OK," to understand. Here, the user's grasp of the pattern or virtual situation which links recent experience with proximate expectations is aligned with the actual flow of state traces through the computational process if the loop is in lock. This is how it works when it works. The mental model of the immediate process context predicts expected responses to user actions and the actual system response to actual user actions resembles this expectation. The user's concept of where they are in the process is continually, and continuously, updated, as is the phase alignment in a phase-locked loop. Surprises constitute discontinuous updates. Here the user feels obliged to "return to square one" and re-establish context, i.e. grok lock. The point of calling it a grok-locked loop is to emphasize that the critical loop closure happens within the user's comprehension. It is the user's construction of the proximate process context (including projection into the future) where success or failure of "the system's response to user input should be predictable" is measured. Each response should be informative. It should bring new information. But it should fit into a pattern of continuity that the user can grasp, retain, and re-apply.

In particular in web browsing, the control actions are very simple and similar, but they all lead different places in the web. The browser loads a new page based on the URL embedded in the hot link. But the choice of which page to load is made based on the user's comprehension of the native content of the page. [Let us use the term native content to refer to the content of a markup-language document exclusive of the markup itself.] The natural language interpretation of the native content leads to user choice of the next automated action (link following) and the link following leads to more native content for the human to interpret.

It is widely recognized that the Graphical User Interface has gained widespread acceptance because it does not require the user to recall large numbers of cryptic commands, but rather can present menu entries and other hints that appeal to the user's recognition memory, which controls a much larger vocabulary than their recall memory. However. in order to complete the cycle of continuing interaction, the situation evoked by the sequence of displays does have to connect with the user's recall memory to find a pattern which extends past what has already happened into the future, after actions that the user hasn't taken yet. This is so the user can anticipate or predict system responses to the available user action options. This can be short-term recall for patterns built up by a sequence of similar responses. It does not always have to be from long-term recall. On the other hand, in situations like web browsing, the user can often start at many places within a structured resource, so it is not safe to count on having approached the current interaction by a particular path.

The peephole effect. People with visual impairments, and some people with certain cognitive functional impairments, do not get the full random-access presentation of the GUI display, however. When the content of a page is read through Text-to-Speech technology, the ideas are presented serially, with approximately a phrase being perceived at a time. This is a very small window on the world that the computer dialog is trying to create. The user's grasp of the context is essentially limited to what they can conceptualize. Here we are back to the recall memory. This is why content is most effective in eyes-free browsing if the exposition of the content is set up on a conceptual grid built on very simply rhythms which are very consistently followed.

Just as text is the pre-eminent data encoding that can be presented to multiple senses, the table of contents or topic tree is preeminent among regular structures to use in gridding your information if you want eyes-free users to be able to move around in it without having to look for guidance. At times the content demands another structure, such as a rectangular array of cells in two or more dimensions. In this case, the dimensions of the array should be commonplaces of popular knowledge, like the days of the week; because the user has to lock the idea of what the coordinate are in their memory where it will lie just below the surface of paying attention to the contents of the cells, but must be easy to recall when time comes to move about.

Plastic Interaction Environments

There is a very general way in which the Grid User Environment can accelerate progress in Disability Access to information, if it indulges in a bit of Universal Design. This has to do with the plasticity of the interaction medium. How easy is it to make changes both large and small to how the computation on the grid is mediated to the human user? The more the Grid pursues agile computation, sometimes referred to as computational steering, and extends this to real-time redirection of the visualization and control channels, the closer the standard computation platform will be to delivering the capability to transform gracefully to make any needed disability accommodations, out of the box.

Articulate Resources

The other dimension of Grid quality that will have a direct bearing on the ease and ability of constructing alternate interfaces is the extent to which the online interface to grid services and resources is self-explanatory. This includes both of two measures: the exhaustiveness of the characterization, and the automatability of the response to this characterization. Articulate resources will capture in their self-documentation everything that is known about the resource. Likewise they will express everything that is so known and documented in the most machine-intelligible form possible.

Opportunities in Grid Forum areas

User Environments

The role of User Environments in The Grid

The Grid is a platform for distributed computation. Computation is what The Grid has to sell. But Computation is not the business that The Grid is in. The cycle of success that makes Grid Computation add value requires success in more than computation. The cycle involves three domains, as shown in the following figure.

Three Domains Illustration [D]

Grid Computation adds value because there are users interacting with applications that are computing on the Grid platform. In considering software architectures for The Grid, it is critical to think in terms of this cycle and not the Grid alone. If advances in the distributability and portability of computation are not well coupled with advances in domain specific application resource pools, and an expanding universe of human-computer interaction environments, the market for Grid services will saturate prematurely and The Grid will not come close to achieving its full potential.

It is not the responsibility of The Grid to create facile and friendly Problem Solving Environments per domain for many computationally-intensive domains of activity. But if the architecture of The Grid does not contribute materially to the integration of such environments by third parties, The Grid will fall far short of its highest and best use.

The same is true on the interaction front. Most human interaction with Grid computation will be mediated at the human through layers of technology which are not Grid specific. It is becoming clear that one of the requirements on Grid generics is that they be usable without any Grid-peculiar capabilities embedded in the user's immediate computing platform. But by the same token, it takes engineering to make resources or services usable from minimal user devices and still take advantage of the full capabilities of immersive theaters. This sort of ruggedizing is a critical Grid architecture requirement.

In particular, one of the challenges for service definition and programming practice for the grid is attaining and maintaining a rich mix and match capability across the three domains sketched above. In visualizing the results of scientific simulation, for example, the display is based on patterns and idioms of the application domain, but populated with specific values from the current computation. The virtual reality is a world whose schema is from the application domain, but whose particulars are from the current computation. That is the kind of mix and match that Grid programming must support to reach the full possibilities of Grid usefulness (and hence utilization).

SMIL and constructive collaboration

Does SMIL have what it takes to do the Grid Job? It would seem to be a good way for a portal to wrap multiple streams of information in a distributed collaboration environment. Only the Grid Forum will know. The technology is entering its second generation with SMIL 2.0 in Last Call. Check it out. Try it out. Show that it works for you or stick it in the W3C ear why it doesn't cope.

X-Forms and scalable control

The work on XML Forms attempts to clean up the ability to create interactive applications in hypertext markup language a bit from what is available in HTML today. It is heavily influenced by a desire to be usable from mobile devices. Has the baby been thrown out with the bathwater? Is it truly industrial strength, or will it break when put to Grid use. This is where the feather edge of the Internet is growing. Be involved, or let others decide what you will have to live with.

Events and mobile messaging

The Grid is going to make it easy for a unique researcher to have a computation undertaken for her that spans the globe and runs round the clock. When there is a problem, there is only one person who has a grasp of the whole computation. Under these circumstances of being on call 24/7 for your running computation, it makes sense to ask that mobile messaging services be used to deliver interrupts to the investigator when exceptions are raised calling for manual review.

Trouble alerts gatewayed to your beeper are but the tip of the iceberg. While people will access the Grid from Grid-specific facilities, there will be many more transactions conducted with people reached over commodity networks and client devices. This requires a continual drive to port Grid User Environment functionality to technology available independent of the Grid.

Implement accessible services

One of the likely products in the Grid User Environments tier will be generic resources for interface building. Because these will be used to gain access to all the other services, and they will be reused over and over, they merit the care to deliver an accessible product.

Contribute to HCI understanding

One of the tasks that still lies before us is creating a workable model of the diversity of user capabilities that service providers can work to. Services such as portals will need to insulate individual service sources from re-inventing their service interfaces for individual users with special needs. The extensibility of Grid service-delivery patterns and the global access to exceptional capability will aid portals or other broker activities in providing this buffering. But the original services will still need to provide some basic flexibility and a clear model of how to access it. This is the transformed, aggregated model of user demand that the middleware forwards on to the original service providers.

There is some knowledge in this area but there is more to be developed. There will be a lot learned in this area in the next few years, coming from many quarters. Participants in the Grid will have experiences that others don't, and will want to learn at the same time from others. Collaboration in this area would appear to be in order. See also the discussion of multilevel description under information services.

Make the Grid work for you

Advanced User Interface capabilities are one area where access to Grid computation can be an asset. Interface prototyping and steering can be realized more quickly with less programming effort if computational capacity is not an issue. Applications of Grid computation in this area can advance the state of science in the area of Human/Computer Interaction. Pushing the limits of User Interface flexibility and agility will also provide the necessary laboratory within which to evaluate Universal Design issues in HCI for Grid computation.

Standard Profile for access to User Services

The Grid way to access services such as help desk and remote interpretation should not have to be invented by each user for their work environment. Portable components that secure access to basic Grid services will pave the way to an ultra-responsive world for users with assistance on demand. What you need­, when you need it.

Abstracting the user, AND the service sources

One of the major roles of the Grid User Environments tier in the Grid architecture is to abstract users away from the rest of the services. The other services have to support a defined range of flexibility, but not random unique requests. This can be a win-win deal for user and backoffice service offeror, but it does not come for free for either. The User Environments tier has to take a reduced set of requirements deeper into the Grid where, if the services feeding the User Environment services do not meet these, the users will not use the service and the service will not have customers.

A major question in this regard is how much abstraction or knowledge the User Environments tier can negotiate from the deeper tiers. People are accustomed to working in a WYSIWYG environment and hard-binding their content to styles. This won't work to reach people with portable devices or Braille displays. It won't look good in composite laboratory command consoles such as one would contemplate building with portal architectures like OPIE. The integrator using a portal to compose a computation control console really want the services to come forward to the portal marked up with Grid-understood html:class markings, for example, rather than random back-office decided styles. Whether the content sources will value the portal's ability to connect them with partner services and people in many diverse equipment environments enough to give up defining the exact look of what they provide is an interesting question. In any case, the accessibility interest is rooting for the Grid to define some functional abstraction conventions such as a class, and let the User Environments tier bind this to appropriate pixels, Braille signs or sound as the user and their situation may need or wish.

Support cross-Grid debug

One candidate service layer for coordinated design across Grid service tiers is the functionality needed for high-level (PSE level) diagnosis or debugging of distributed computations, across the Grid. PSE-level debugging means that information at the lowest level of abstraction needed to isolate and diagnose the problem is blended with context based on the highest-level results presentation possible. For an example from the realm of Web accessibility diagnosis and repair, consider the WAVE tool from Temple University.

User Services

Drive the Grid Quality Bus

Don't let mishaps be forgotten until the alternative is consistent. The second time it happens, it becomes a type. The third time, a problem requiring remedy.

Make the Grid Work for You

The Grid has to be on a track to delivering much more service with somewhat less pampering (human effort put into support). Demand the level of support so that you can collect enough information to show how that can be done.

Create systemic responsiveness

Be the FEMA of the Grid; never get caught without a contingency plan. Any place that booking and using Grid service requires human intervention should be on the User Services map. This map should capture the operational procedures layer of the Grid across the board.

Interoperable trouble event tracking

One of the first requirements to put us on the path to reliable cross-Grid delivery of computational service is a Grid-interoperable record of trouble events.

Special needs

The User Services community in the Grid is a likely candidate to organize centers of special expertise such as Grid services accommodation for people with disabilities.

Compare/share methods with AT support activities

Diagnosing and remedying integration problems between Assistive Technology and mass-market platforms that it has to interoperate with is difficult. Two vendors must bring their expertise to bear at once to resolve the problem. However, users of Grid applications are likewise thinly spread and using combinations that not many have used before them. The requirements for Grid User support and for Assistive Technology support are similar. Compare notes and both will likely benefit.


As discussed above under the portable point of sale, many people with multiple and severe disabilities have custom-configured human:computer interface arrangements and cannot use standard I/O devices. But they can get a lot done if allowed to work within the user environment that works for them. A key need as we move into an electronically-integrated world with pervasive intelligence, is that when transactions need a signature, for example, that the merchant be tooled up to accept a digital signature applied using the customer's digital assistant.

The Grid is going to have a need for the capability to establish and merit trust while communicating over open networks. The communication pattern is not necessarily predictable far in advance. The basic capability to play a trusted role has to be in the playbook that all Grid players follow. The Grid methods of setting up trust communities should be highly robust in terms of going to unexpected places to obtain computational services. The experience of the Grid in this respect can help establish the evidence to show that commerce can and hence must be conducted on a similarly open security platform.

Information Services

The following scenario relates to the representation of Grid offerings of computational service. This is a topic of joint interest to Information Services and to Scheduling.

Discovery Plus

Here is a rough overview sketch of the role of information in the process whereby a Grid user books and receives Grid services.

Who can do my work?

The first round is the discovery of resources. The user has an idea for a computation that they which to perform. Let us presume it is a novel experiment in computational science that nobody has ever done before. So the use does not have a habitual supplier or a compute factory in mind. The first step is to roughly break the proposed task into services which when combined would accomplish the entire experiment. Then the user goes shopping for resources. Their basic question at this point is "Who can do my work?" The definition of "my work" is at a reasonably high level, mentioning only the few leading performance parameters of each piece, which would suggest that the user may have trouble finding suitable services. The objective is to have a short list of qualified sources for each role in the map of service utilization, and then not too many more. But this search resembles a typical web search. What the user says about what they are looking for is a few characteristics drawn from a relatively large dictionary of properties. The user may loosen or tighten this dragnet to get to a compact list of candidate sources.


At this point it comes time to talk turkey. Exactly just what service does each offeror offer? What does it take to book and buy it? This will include very low-level data that are required to ensure successful and efficient interoperation with the other service providers, possibly at a distance. The offers and demands of all the service offerings are compared along the path of all service flows, and the best mutually-supported profile of tuning parameters is elected for various candidate service constellations: graphs of linked service sources.


Then the overall performance profile of the leading few candidate service constellations in dollar cost, calendar availability, risk factors, etc. is compared. Then reservations for the services in the winning candidate are made or confirmed and the work is in the queue.

Big schema, small catalogs

The apparent moral of this tale is that there is a large dictionary of interoperation-critical parameters that needs to be understood grid-wide, but that most of the detail enters the process in the form of direct communication between a candidate service provider and a prospective customer, after candidate service providers have been identified. The details of the final compatibility check and optimization of the service exchange do not particularly have to be published and distributed into global discovery services, just the properties that make a service exceptional and competitive worldwide. The customer does not want to address that level of detail in the initial lookup. They wish to assure some range of options returned from the initial search. The compatibility requirements often come from other discovered services in any case, so the user does not know all of them in the discovery phase.

Multilevel description

The above conclusion suggests that the description of service offerings spans multiple levels of abstraction, a high level for competitive identification of capable resources, and a lower level for cooperative confirmation of compatibility and optimization of efficiency within a concrete plan for performance of the computation.

The high level, particularly if supported with a range of options in the details, opens wide the maximum market for a given service. This also fosters the ability to do unusual tasks as could be required to complete a pattern of service with a node performing real-time translation or extraction to haptic model information on the fly. This level of flexibility is in the service offering itself and not created by the way the service is publicized. However, service offerors will wish to offer flexible services to the point that it is only the high-level performance properties make sense to mention until some details of the actual work to be done are known.

Interest in multilevel abstraction in the description and subsequent negotiation of services is active in various quarters in the larger world of Information Technology. The CONNEG working group in the IETF, the CC/PP work in the W3C, and the AIAP project in NCITS/V2 are all representative of current interest in enhancing functional capabilities in this direction. Capabilities such as these would appear to hold considerable promise in making graceful transformation of user interfaces to meet the needs of people with disabilities...

Table of Contents:

Valid HTML 4.0!