SVG images are uncommon compared to png, gif and jpeg so when we implemented support for them we wanted to keep to the mental model that people have for raster images.
If SVG images work just like raster images then websites that support raster image upload can allow SVG images to be uploaded too without having to worry about implementing additional security or privacy checks.
So let’s see what we expect from raster images:
- Comprise a single file
- Are not interactive i.e. for example they can’t react to you moving the mouse over them
- Should display pretty much the same independent of the operating system you’re running
- Cannot change to display something else depending on the time they are viewed
- Behave like a raster image to the containing document i.e. they don’t have a DOM
- Cannot have areas that work as hyperlinks
- Will not render differently depending on whether particular sites have been visited
- Cannot send or receive data
- But they can be animated
If we omitted one of these restrictions, allowing hyperlinks to be displayed in visited colours for example, copying the image to canvas via drawImage would then allow a web page to determine whether another site had been visited by the user unless we made the canvas write-only. As we don’t make the canvas write-only when a raster image is copied to it, we don’t want to make it write-only when an SVG image is copied to it either.
So if you’re going to use SVG as a background-image, border-image or via a html <img> or SVG <image> element and your image needs to do something that a raster image could not do then it’s unlikely that will work for you. Having the image on the same domain as the host document won’t make any difference.
Some workarounds for common problems:
- If you want to use a stylesheet, you can encode the stylesheet contents as a data URI in the image file
- If you want your SVG image to contain other images then you can encode the embedded images as data URIs
- While hyperlinks in images are disabled, you can make the whole image a hyperlink by wrapping a hyperlink tag round the image element itself in the container document
- If you need anything that images don’t offer then perhaps you should be considering putting the SVG inline in the document or using an <iframe> or <object> tag instead
Firefox has supported all SVG 1.1 filter primitives since Firefox 3, however until now it has only supported SourceGraphic and SourceAlpha pseudo inputs. Fortunately these are by far the most common.
Firefox 17 includes support for an additional two pseudo inputs – FillPaint and StrokePaint. The SVG testsuite has an example showing how they can be used.
Even though HTML doesn’t use the SVG fill and stroke CSS properties, you can still set them in Firefox to have StrokePaint and FillPaint filters with HTML.
An SVG feature appearing for the first time in Firefox 15 is support for SVG Fragment Identifier linking. This feature allows authors to display different parts of an SVG image by specifying a different fragment identifier on the end of the URL.
The Firefox 15 implementation also includes support for the view element which is basically a predefined fragment identifier. It does not however, support XPointer style fragment identifiers.
One use case for fragment identifers could be an SVG document with a set of sprites arranged in a grid. Fragment identifiers could then be used to display one or more of those images on another html or svg page.
Here’s an example that shows Fragment Identifier Linking in action
One of several new SVG capabilities in Firefox 15 is support for triggering animations by clicking on hyperlinks.
If a link target is an animation element then the animation will begin when the user clicks on the link such as in this example.
This completes SVG 1.1 animation support in Firefox, apart from the deprecated animateColor element and wallclock timing, both of which are likely to be removed in SVG 2.0. Therefore from Firefox 15 we’re also enabling the http://www.w3.org/TR/SVG11/feature#Animation feature string.
Support for the SVGTests interface has just landed in Firefox. It’s currently available in Nightlies and all being well, should make its release appearance in Firefox 12.
The SVGTests interface allows scripts to check whether additional functionality is supported by a browser. There’s no strict definition of what extensions mean but Firefox currently uses element.hasExtension(“http://www.w3.org/1999/xhtml”) to suggest that foreignObject elements support html content and element.hasExtension(“http://www.w3.org/1998/Math/MathML”) to suggest they also display mathML.
The SVGTests interface also allows the requiredFeatures, requiredExtensions and systemLanguage attributes to be manipulated as if they were arrays using the SVG DOM.
According to the original SVG 1.1 specification masks in SVG always operated in the linear RGB colour space. That meant that Firefox always converted the graphic being masked to linear RGB and after that, applied the mask.
SVG 1.1 2nd edition changes the way masks work and a patch has landed for Firefox 10 to match that change. SVG masks now honour the color-interpolation property. This means that authors can choose whether the graphic is converted to linear RGB before the mask is applied.
The thing to watch out for though is that the default value for color-interpolation is sRGB so existing content may render slightly differently. If you want things to stay the same you need to add color-interpolation=”linearRGB” as an attribute of the mask element.
Opera has worked like this for some time now and a patch make this change also landed in Webkit recently so Chrome and Safari will render things this way at some point too.
In versions of Firefox prior to 4.0 there was an svg.enabled flag that you could set to false in about:config to disable Firefox’s SVG capability.
During the development of Firefox 4, UI changes took more and more advantage of SVG; for buttons for instance you can ship with fewer bitmaps – one scalable drawing can replace all the bitmaps for different screen resolutions and using SVG filters you can even derive the greyscale disabled state from the enabled button.
Eventually we discovered that the Firefox 4 UI had become so internally dependent on SVG that it would not start any more when you set svg.enabled to false so we removed the
flagfootgun. SVG is now a first class citizen just like html.
One consequence of this is that if you were using an SVG plugin such as the Adobe or Corel SVG viewers these will no longer function. To ease the pain, we have implemented more of the SVG specification in Firefox 4 than ever before and as you can see we’re up to a similar score to the Adobe plugin. There are still some things that the Adobe plugin does that Firefox does not, such as SVG fonts there are now things that Firefox does or does correctly that may in some way make up for that.