A good cache system for temporary files really needs to be implemented at the system level, not at the application level. This way the system can clean up temporary files of all types from all applications when space is running low. The system should use all of the remaining unused space available for temporary files. I wasn't kidding when I created this image of cache settings. To ensure that users don't get worried about low disk space, the system should ignore the cache when calculating the total free disk space. For instance, if normal files are occupying 60% of available space, and temporary files 38%, the system will still report that there is 60% of the space free. Actually, memory caching should be done the same way.
10:25 PM New Work
I've accepted a contract position with Mozdev Group to work on some Mozilla-related projects. Actually, I started on Monday so I've already done some work already. It's always good to work with folks who share the same vision as I do. Let's hope together we can build some exciting products and help to promote Mozilla technologies.
5:28 PM Building Technologies with Baby Steps
When someone encounters somthing new, one has a tendency to compare it to something that already exists. This is a natural habit that everyone does to help provide a starting point for determining how to react to whatever this new thing is. For instance, when a new product is released, it will be compared to some existing product, and lists of advantages and disadvantages between the new and old will be created. It really doesn't matter that the new product has a very different purpose or not.
For instance, if you build a new revolutionary killer product that does all kinds of never-before-done things, but you put a Back button, and an address bar and Bookmarks on it, it is no longer a new revolutionary killer product, but just another web browser. It doesn't matter what you say, because people will look for the familiar in your new product, and compare your product to that. You can't create revolutionary products intentionally, so quit trying.
Why was XML successful? Likely much of its success was due its similarity to HTML. If XML had used square brackets or other different syntax, its possible that it wouldn't have gotten as far. People need a way to transition to new things so that familiarity is maintained thoughout the process. When building a new technology, you need to ensure that radical changes are minimal to encourage adoption.
For instance, a web developer that understands HTML forms would have no trouble with a few new tags or attributes added to HTML forms, but would really get lost when it comes to XForms. That would require one to grasp XML syntax, different tags for forms, the separation of UI and data, XPath, and a number of other things. The intuitive leap is just too great. Instead, technologies must be created which allow the developer to take baby steps from one concept to the other. In fact, Opera has provided an alternative to XForms which is much more transitional, and, if suitably implemented, would likely be more widely adopted.
Have you noticed how popular the web is? The web has slowly changed the way that developers and companies think about things. You can't just light a match and blow it away. Progress is done in small steps.
7:26 PM One more XAML thought
One thing I noted about XAML is that both the code and the XML parts need to be compiled before using them -- although the XAML can be interpreted by Explorer in some means, perhaps by compiling it when needed.
From what I can tell there is no XML at run-time, meaning no DOM-manipulation. You need to use all new APIs to manipulate the UI. Actually, the whole thing works very similar to Mac OS X, which stores UI descriptions in XML which map to classes. Apple doesn't promote this much though as they prefer that people use the visual tools.
6:38 PM Some XAML comments
OK, I took a quick look at XAML to see how it compares to XUL, or anything.
- It is indeed a XUL-like language for describing user interfaces. Why didn't they just use XUL? Well, you wouldn't expect MS to use something someone else created would you?
- XAML might be only a codename. Sometimes, the documentation says it's a codename, and XAML always appears in quotes.
- You can run compiled code in a standalone window or in a browser, but you can't do both. Switching requires recompiling.
- Each 'XAML' tag corresponds to a Class. Attributes on the tags correspond to the fields (properties) of the class. For example, a Button class can be specified with <Button Width="100" FontSize="10">
- There is no CSS used -- instead specific attributes are used. There is something confusingly called Styles, which appear to be more like CSS classes for sharing appearance with a number of elements. This Styles feature does seem to have the capability to adjust the appearance of inner content, for example the thumbs and slider part of a scrollbar can be adjusted with a block of XML. In fact, some of the examples suggest that the Style features might have some XBL-like content capability, although it looked more confusing.
- There's a databinding feature, kind of like the one IE already has which can bind to XML, SQL and so forth. In XUL, this is done with templates and RDF. In XAML, the data to bind to are even called data sources.
- There's something which falls into the category of 'Yet another XML vector language which isn't SVG.' It does look very similar though, but my cursory glance and limited knowledge of SVG suggest that it isn't the same. It is called Windows Vector Graphics though. Yes, this is exactly what the world needs. More similar but different vector graphics languages.
- As far as I can tell, there's no XBL or overlays or anything similar. Custom components can be created in native code (C# or whatever). So, there's isn't anything as cool as Firebird Extensions.
- The documentation is, like most of the stuff on MSDN, difficult to navigate through. Most of it is vague and more of a reference. There is only a few pages that actually explain how to do something. Because of this, there may be something I'm missing. There may be more interesting features that I missed, but I don't think XAML itself is necessarily all that powerful. In fact, it's really just a mapping between classes and XML.
There's a general overview here.
5:21 PM XAML Info now available
XAML info is now available. I haven't looked at it yet, but it's rumoured to be similar to XUL. More comments to come soon.
1:09 PM XUL-related quotes from Hixie
Ian Hickson has some interesting quotes from a recent XUL naming debate between him and, if you've been following XUL, you should know who else.
We stopped working on the XUL spec itself when XUL Planet started documenting XUL in more detail than we had time to do ourselves.
Most of the current XUL work is being done within standards organisations with strict NDA policies. If you are a ISO, W3C, ECMA, or IETF member let me know and I can show you the relevant links... I'm under NDA from one of the above groups, in which XUL is being discussed.
XUL _is_ in active development, we are currently taking feedback into account and are more carefully defining the XUL box model.
All quotes from here.
7:14 PM Custom XUL Rendering Widgets
I've been thinking a bit over the last week or so about support for custom drawing widgets in Mozilla. There are four possibilities:
- Create a <canvas> tag as described in bug 102285. Although the patch needs some work, this is probably the easiest approach. However, it limits one to a single special XUL tag for drawing.
- Create a modifiable image object. The image would have a variety of drawing functions such as drawLine, drawRect and so forth as well as pixel-level functions. It might even be possible to support transparency and animation as existing images do. It could then be used in the HTML and XUL image tags, or even CSS properties if a method was created to assign custom images to them. The canvas tag patch suggests that this technique would be quite simple to implement. It has the advantage that it could work on various different tags.
- Create an interface with drawing functions. These would be called by the application during paint events (such as onpaint) to update the display. This would have the advantage of not necessarily requiring a buffer to hold image data, but would be slower for repeated redraws. It would however, allow extra components to implement the drawing interface, allowing for various alternate system-specific libraries to handle drawing.
- Use SVG. This is the most work, but only really handles some cases. SVG requires a DOM node for every part of what is to be drawn, which isn't practical for pixel-level drawing or heavy animation.
There's also some security issues, for example, to ensure that unprivileged code either can't use custom drawing, or can't draw over things it shouldn't be able to.