- From: Thomas Roessler <tlr@w3.org>
- Date: Wed, 28 Mar 2007 19:06:28 +0200
- To: public-wsc-wg@w3.org
I forwarded (with permission) the message below to Chuck in discharging ACTION-161, but forgot to copy the list. -- Thomas Roessler, W3C <tlr@w3.org> +33.4.89063488 ----- Forwarded message from "Ian B. Jacobs" <ij@w3.org> ----- From: "Ian B. Jacobs" <ij@w3.org> To: Thomas Roessler <tlr@w3.org> Cc: mike@w3.org, dom@w3.org, shadi@w3.org, cooper@w3.org, dsr@w3.org, w3t-tech@w3.org Date: Sun, 11 Mar 2007 19:54:29 +0000 Subject: Re: Defining different user agent classes for conformance? X-Spam-Level: Organization: W3C X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.1.5 On Fri, 2007-03-09 at 23:50 +0100, Thomas Roessler wrote: > The Web Security Context Working Group will make recommendations > about usable presentation of certain information (and on trusted > paths to getting to that information). In making such > recommendations, will need to make certain assumptions about the > user agents that it is talking about. (E.g., there is a viewport, > there is just an acustic interface, there's very limited screen real > estate, there is no keyboard, ...) > > I'm wondering what the prior art is in terms of defining classes of > conforming implementations in terms of the interactions and > capabilities of clients. > > Mike, Dom, Shadi, Michael, Dave, any insights that might help? Hi Thomas, We wrestled with a similar question in UAAG 1.0. In the end, the conformance approach we adopted [1] works like this: * UAAG 1.0 is a list of checkpoints. * We chose not to allow people to conform to arbitrary subsets; that would not provide a sufficient degree of interop. We felt that the UAWG understood best which sets of checkpoints needed to be satisfied as a group. * Thus, we provided some conformance options by grouping related checkpoints into labeled (indivisible) subsets. For example, the "video" conformance label corresponds to a set of checkpoints. * People then pick and choose which conformance labels apply to their user agent. There is a core set that every user agent must conform to. After that, people choose the conformance profile that most suits their needs. * In any conformance claim, you have to indicate what conformance profile you used. This approach thus provides some flexibility, but tries to ensure that coherent sets of requirements are always implemented together before one can claim conformance in that area. Some sets of checkpoints identified in UAAG 1.0: * A, AA, AAA * Image, Video, Animation, Audio, Speech, VisualText [Note, for example, that someone claiming conformance for an audio/speech-only browser would thus not have to conform to the Image, Video, Animation, and VisualText sets.] * Events (i.e., whether input-device eventing is supported). * Selection (i.e., whether the user agent supports the "selection" abstraction). * Pointer (i.e., mouse), Voice. [By default, all user agents are required to support keyboard input and may also conform to these two other input device types.] You may find labeled sets of requirements provides some conformance flexibility. Your groups of requirements may be tied to more concrete device capabilities than those of UAAG 1.0. _ Ian [1] http://www.w3.org/TR/UAAG10/conformance#Conformance [2] http://www.w3.org/TR/UAAG10/conformance.html#conformance-profiles [3] http://www.w3.org/TR/UAAG10/conformance.html#conformance-claims -- Ian Jacobs (ij@w3.org) http://www.w3.org/People/Jacobs Tel: +1 718 260-9447 ----- End forwarded message ----- ----- Forwarded message from "Ian B. Jacobs" <ij@w3.org> ----- From: "Ian B. Jacobs" <ij@w3.org> To: Thomas Roessler <tlr@w3.org> Cc: mike@w3.org, dom@w3.org, shadi@w3.org, cooper@w3.org, dsr@w3.org, w3t-tech@w3.org Date: Sun, 11 Mar 2007 14:58:44 -0500 Subject: Re: Defining different user agent classes for conformance? X-Spam-Level: Organization: W3C X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.1.5 On Sun, 2007-03-11 at 19:54 +0000, Ian B. Jacobs wrote: > On Fri, 2007-03-09 at 23:50 +0100, Thomas Roessler wrote: > > The Web Security Context Working Group will make recommendations > > about usable presentation of certain information (and on trusted > > paths to getting to that information). In making such > > recommendations, will need to make certain assumptions about the > > user agents that it is talking about. (E.g., there is a viewport, > > there is just an acustic interface, there's very limited screen real > > estate, there is no keyboard, ...) > > > > I'm wondering what the prior art is in terms of defining classes of > > conforming implementations in terms of the interactions and > > capabilities of clients. > > > > Mike, Dom, Shadi, Michael, Dave, any insights that might help? > > Hi Thomas, > > We wrestled with a similar question in UAAG 1.0. In the end, > the conformance approach we adopted [1] works like this: > > * UAAG 1.0 is a list of checkpoints. > * We chose not to allow people to conform to arbitrary subsets; > that would not provide a sufficient degree of interop. We felt that > the UAWG understood best which sets of checkpoints needed to > be satisfied as a group. > * Thus, we provided some conformance options by grouping > related checkpoints into labeled (indivisible) subsets. For example, > the "video" conformance label corresponds to a set of checkpoints. > * People then pick and choose which conformance labels apply to their > user agent. There is a core set that every user agent must conform > to. After that, people choose the conformance profile that most > suits their needs. > * In any conformance claim, you have to indicate what conformance > profile you used. > > This approach thus provides some flexibility, but tries to ensure that > coherent sets of requirements are always implemented together before > one can claim conformance in that area. > > Some sets of checkpoints identified in UAAG 1.0: > * A, AA, AAA > * Image, Video, Animation, Audio, Speech, VisualText > [Note, for example, that someone claiming conformance for > an audio/speech-only browser would thus not have to > conform to the Image, Video, Animation, and VisualText > sets.] > * Events (i.e., whether input-device eventing is supported). > * Selection (i.e., whether the user agent supports the "selection" > abstraction). > * Pointer (i.e., mouse), Voice. [By default, all user agents > are required to support keyboard input and may also conform > to these two other input device types.] > > You may find labeled sets of requirements provides some conformance > flexibility. Your groups of requirements may be tied to more > concrete device capabilities than those of UAAG 1.0. Whoops, I left out the "tag line": UAAG 1.0 does not define "user agent classes"; that proved too challenging. We did not find we were able to define "a visual browser" or "a voice browser". One can often extend a browser with a plug-in, making it difficult/overly constraining to define rigid classes. The UAAG 1.0 approach is to define conformance profiles, therefore, and let people piece together the profile that best corresponds to their user agent's capabilities (at a given moment) and to which they wish to claim conformance. _Ian > [1] http://www.w3.org/TR/UAAG10/conformance#Conformance > [2] http://www.w3.org/TR/UAAG10/conformance.html#conformance-profiles > [3] http://www.w3.org/TR/UAAG10/conformance.html#conformance-claims -- Ian Jacobs (ij@w3.org) http://www.w3.org/People/Jacobs Tel: +1 718 260-9447 ----- End forwarded message -----
Received on Wednesday, 28 March 2007 17:06:31 UTC