- From: David Bolter <david.bolter@utoronto.ca>
- Date: Tue, 25 Mar 2008 10:03:54 -0400
- To: Henri Sivonen <hsivonen@iki.fi>
- CC: public-pfwg-comments@w3.org, Ian Hickson <ian@hixie.ch>, Maciej Stachowiak <mjs@apple.com>, david.bolter@gmail.com, wai-xtech@w3.org
Hi Henri, Thanks for expanding on your points. The "black box" argument is worrisome and I understand your point much better now. My mind is stuck in the open source world and I tend to think in that realm. The incremental innovation argument also makes a lot of sense. Given the potential import of keeping or removing this aria property I'm cc'ing wai-xtech in case we are missing something. cheers, David Henri Sivonen wrote: > Hi, > > On Mar 24, 2008, at 21:35, David Bolter wrote: >> Henri Sivonen wrote: >>> According to >>> http://mindforks.blogspot.com/2008/03/aria-templateid-explained.html : >>>> Whenever an assistive technology (AT) sees this template ID it can >>>> provide customization to improve the UX. For example, a screen >>>> reader like JAWS or Orca could load a script for adding keystrokes >>>> to open the Gmail inbox etc. If another web product embeds Gmail >>>> the AT can still pick up the the template id and apply some >>>> customization. >>> >>> >>> I think this is a problem for innovation and for competition. >>> >>> It is a problem for competition, because if JAWS or Orca give magic >>> preferential treatment to Gmail, a competing Webmail service cannot >>> get the same accessibility advantage by writing to an open spec. The >>> playing field is not level in that case. The only way for the >>> competing service to gain the same advantage is to reverse engineer >>> what the template ID causes JAWS or Orca to do and then carefully >>> craft their own UI to match the structure that the magic behavior >>> expects. What's the point of an open spec like ARIA in that case? >>> Moreover, such crafting might be too impractical for entire UIs like >>> a Webmail UI. The reverse engineering scenario could be a realistic >>> recourse for an Ajax library when the template ID is associated with >>> a specific widget of a competing Ajax library as opposed to the >>> whole UI of an app. In either case, the anti-competitive effect is >>> not good. >>> >> I think you are suggesting that anti-competitive practice is a >> potential abuse of this ARIA property? > > I'm suggesting that aria-templateid inherently has anti-competitive > consequences. > > If site/library A gets preferential treatment triggered on its > template ID, site/library B cannot get the same treatment by writing > to any public spec/API. Worse, since the preferential treatment isn't > even a private API but arbitrarily complex behavior triggered > potentially inside a black box, it may not even be feasible for > site/library B to reverse engineer what is going on and copy the > template ID. > >> I think it is good to consider such things. I do think that the >> ability for someone to better customize his/her AT is empowering and >> perhaps worth the gamble? Consider if we already had the ability to >> do this in the web or desktop space, would you want to take it away? >> JAWS scripters have clearly felt the need to do this on a per URL or >> per desktop application basis. > > We do have Greasemonkey scripts already. They are different though. If > there's a Greasemonkey script in circulation for site A, site B can > see which APIs that script calls and even make site B use those APIs > by default. Also, I'm not really that worried about users circulating > AT scripts that others can read like they can read Greasemonkey > scripts. I'm concerned about special behavior inside the AT black box > triggered by a magic key. > > As for desktop apps, building app-specific knowledge/hacks into AT > would be anti-competitive there, too. Suppose popular AT has private > magic enhancements for office suite A. Office suite B cannot gain the > same functionality by writing to a pre-existing public API and may > have to work closely with AT providers to get their app supported. > >>> It is a problem for innovation, because Gmail is getting >>> preferential treatment for a particular version of its UI, updating >>> the UI in an innovative way would regress accessibility. With the >>> glacial update cycle of AT, it isn't reasonable to expect new AT >>> versions to be shipped instantaneously. Hence, aria-templateid would >>> cause a disincentive to change anything that has once gained >>> preferential treatment from AT. >>> >> I would hope that the UI is accessible/useable with or without the >> customization provided by the AT based on the templateid; that the >> templateid is the way to provide the "icing on the cake". In this >> scenario the UI authors should feel free to innovate and change the >> templateid (eg. "foo.com/gadget1.0" becomes "foo.com/gadget1.1"), >> effectively removing the icing until someone reapplies it. > > Having to remove the icing in order to innovate is a barrier to > innovation. You then can't innovate in small increments whose value > (per each increment) is smaller than the value of the magic icing. > > Consider the upgrade barrier caused by AT working with one version of > an office suite but not the next on desktop. The problem becomes worse > with the Web 2.0 release cycle. >
Received on Tuesday, 25 March 2008 14:04:40 UTC