- From: Maciej Stachowiak <mjs@apple.com>
- Date: Wed, 18 Feb 2009 10:46:48 -0800
- To: David Bolter <david.bolter@utoronto.ca>
- Cc: Steven Faulkner <faulkner.steve@gmail.com>, HTML WG <public-html@w3.org>, W3C WAI-XTECH <wai-xtech@w3.org>, Janina Sajka <janina@rednote.net>
On Feb 18, 2009, at 10:05 AM, David Bolter wrote: > Ben, a Bespin developer has written about some accessibility > thinking here: > > http://benzilla.galbraiths.org/2009/02/18/bespin-and-canvas-part-2/ > > It seems reasonable to me. How do we move this conversation forward? The downsides to a bespin-like approach to text editing go beyond accessibility. Operating systems provide quite a few built-in services for text, and it's hard to recapture them when doing raw text drawing onto a bitmap surface. I tend to draw parallels to native APIs in cases like this. <canvas> is a raw drawing area. Most OS-level equivalents let you use additional low-level APIs to hook into accessibility, international text input, spellchecking and other system services. It makes sense to me to have some of these, to enable <canvas> to be more effective as a mechanism for bleeding-edge innovation on the Web. However, designing low-level APIs to handle these details in an OS-agnostic way is pretty challenging. This is why my initial reaction is to figure out what is inadequate about current HTML editing facilities for a project like Bespin and add the needed functionality. That's a conceptually simpler problem to tackle than "what are all the OS-level hooks you miss out on by doing your own text drawing and editing". However, it seems to me that part of the reason Bespin was written the way it was is simply that the authors are comfortable with that style of programming and enjoyed writing it that way. Past experience with Java's 2D graphics APIs was cited. Consider that Bespin includes a canvas-rendered UI widget toolkit, and it seems to me that none of the arguments for <canvas>-based custom text editing really apply there. So we may actually need someone else to try to rebuild the Bespin experience in a non-<canvas>-based way and report the stumbling blocks. But I am not sure anyone will find that to be a fun and motivating project. As to the specific proposal there, to have a hidden DOM under the canvas element, I'm not sure that is the best way to expose OS integration, not even just for accessibility. If such an approach works with acceptable accuracy and performance then it casts doubt on the motivation for using <canvas> at all; and in general having two completely separate ways of presenting the same content that must be kept in sync (even in the face of dynamic change) is likely to be error-prone. Regards Maciej
Received on Wednesday, 18 February 2009 18:47:34 UTC