- From: <i@hansschmucker.de>
- Date: Thu, 14 Jul 2016 14:10:13 +0200
- To: <public-html@w3c.org>
Thanks to the feedback from both the W3C and WHATWG lists I think that the gallery element is not necessarily something we should focus on. At least not in the sense that the inline presentation is of much importance. Designing the overview and embedding it is something that can be handled by using existing standards and there are other proposals like panelssets ( https://bkardell.github.io/common-panel/prototype/panelset-element.html ) which will make creating the overview even easier. I'm not saying this is something that we should abandon entirely, but I think it's safe to spin this off into a seperate issue. So let's focus on the remaining issue: While there is an easy method for showing individual images in a way that is appropriate on both desktop and mobile (namely a simple link, which any UA will display appropriately, along with a way to return to the current document), there is no way to do the same for -images with description -series of images (with or without description) -items of any kind which can be rendered as an image The only way people are able to render such data is by either including all images in the gallery directly or as a series of links. The UA on the other hand is unable to produce anything from this series of links without guessing since there is no standardized format for determining if a series of links form a gallery. Microformats ARE able to describe this and in fact a UA could choose to render a series of microformat-described figure elements in a desirable way. The remaining issue for me seems that the interaction between this "viewer" and the page is completely unspecified and that items like descriptions and copyright notices which may be required to make sense of an image may or may not be shown. Also, an image may represent an item of any kind or offer actions on it (for example "Download the PDF", "View the subgallery", "Show related images", "Go to full description"), which the page may be unable to perform if there is no standardized way to interact with the UA. -------------------------- Proposal: Standardize a way to declare a union of items which are represented by an image, have (optional) essential descriptions, can be moved outside the context of the page and can offer additional actions by interacting with the page via url and script. There are a couple of existing standards which already take care of some of the issues, for example FIGURE, PICTURE and IMAGE. Whether other items like VIDEO or CANVAS can adequately be represented is something I'm not entirely sure about. While it is certainly desirable, the burden of implementing this may be too big. There's a pretty good chance that we'll get inconsistent results in that case. For the interaction with the page we may want to take a page from FORMs, with each action essentially being a form action which may be performed via a FORM element or captured using appropriate event handlers. No need to reinvent the wheel. Invoking the gallery is actually a bit of a problem, since we have no directly visible element which can be used for this purpose. I'd suggest making the gallery handle all clicks, while reminding authors that they can simply stop propagation. One major issue that's still open is how we are going to differate between the content meant for rendering the gallery as a whole and the one meant ot be displayed full-size. A UA-rendered gallery overview would have avoided the issue but with custom galleries, there's going to be a lot of content that is really just styling. On the other hand, we don't want to punish developers with clean CSS by forcing them to include the same content twice. A single FIGURE element should be sufficient to describe thumbnail and expanded view. I'd suggest simply making the gallery element very strict about what it accepts as valid content, while still giving a defined behavior for cases when there's "invalid" content. That way "invalid" content can be used for presentation purposes. Based on all this I'd suggest the following: The gallery is a block element containing figure and/or other content. It has no visible representation aside from its content. Each figure which is a direct or indirect descendent of the gallery element represents a frame. The viewer is invoked when a click event that was triggered on a FIGURE element bubbles up to the GALLERY. The UA presents the frame to the by performing the following steps: *The figure element gets a min-width of the full-screen area *The figure element gets a min-height of the full-screen area *The figcaption gets a display:none *All form elements inside the figure element are set to display:none *A focus event is fired on the figure element *A fullscreen event is fired with the fullscreen element set to the figure element *A snapshot of the figure's DOM is taken *The UA takes a snapshot of the figcaption's textContent *The UA compiles a list of all input type=button and input type=submit elements inside the figure *The UA may slow down rendering of the page at this point *The snapshot is rendered to an image (the UA SHOULD choose an appropriate density, at least 2x the display density) *The UA MAY prerender other frames inside the galleries according to the same rules (except it doesn't raise another fullscreen event, but sets the fullscreen element) *The UA MUST update the rendering when new resources become available, for example if a picture has been loaded at a higher resolution *The UA displays the image with platform-appropriate controls, it SHOULD display controls for navigating among the images and MUST display controls for exiting the viewer mode *The UA MUST present the figcaption textContent along with the rendered content (however it may hide it after a brief period) *The UA MUST display the list of INPUT elements. Activating one of these INPUT element closes the viewer and raises the appropriate event on the INPUT element (it MAY however hide the list behind a menu) *When the user navigates to a different figure, the UA MUST raise a focus event on that element before rendering it *On close, a full screen exit event is fired Example HTML <gallery> <div> <figure> <picture><img src="..." /></picture> <figcaption>...</figcaption> <figure> <div> Additional controls <div> </div> <figure> <form action="download.php" method="GET"> <input type="hidden" name="pic" value="..." /> <input type="submit" value="Download" /> </form> <picture><img src="..." /></picture> <figcaption>..</figcaption> </figure> </gallery> Still pretty broad strokes, but I think we're getting there. I'll try to write up an implementation for Firefox desktop and mobile during the weekend, so we can play around with it a little more.
Received on Thursday, 14 July 2016 12:10:51 UTC