Mozilla Foundation and Opera Software Describe Joint Vision for Web Application Framework

Tuesday May 25th, 2004

The Mozilla Foundation and Opera Software have published a paper outlining their vision for Web applications. The paper, submitted in preparation for next week's W3C Workshop on Web Applications and Compound Documents, describes a device-independent Web application framework based on HTML and backwards-compatible with existing Web content. The two organisations are keen to get parts of this framework in place soon to prevent a single-vendor solution (see Microsoft's position paper) becoming dominant. Co-author Ian "Hixie" Hickson's provides more insight in a recent weblog post on the matter.

#14 Re: KISS *is* the right way forward...

by jilles

Wednesday May 26th, 2004 10:00 AM

You are replying to this message

KISS is the primary reason HTML is useless for semantic markup. It's just too limited. CSS was bolted on eventually but suffered from the same and the opposite problem (insanely complicated to implement and you are still severely limited in layout). Now the same mistakes are made for webapplications.

Basically it is a reinvent the wheel exercise. We already have mature, capable client side guitoolkits (win32, gtk, qt) that allow us to build very rich clientside applications. The problem with these toolkits that webapplications try to solve is the software distribution and the security (it's not very efficient and secure to distribute applications like that). Ideally a solution would be equally powerfull as todays client side gui toolkits (why accept less?). The reason we are currently settling for much less is that this apparently is technically hard on top of the w3c set of technologies. This is not surprising because these were never designed to do more than what we are seeing today (much less actually, it is amazing how people work around limitations in existing web technologies). Css is nice until you realize that a lot of features that you take for granted in a dtp package, word or openoffice are simply not there. (Want columns? Forget it.).

The assumption that people don't want/need such features has proven wrong. And consequently CSS, HTML & javascript are totally inadequate for what they are used today. I work for a company that produces a content management system. Maintaining the GUI is a PITA. Basically the GUI would not be accepted by users if it wasn't running in a browser. Compared to our competitors, we actually have a very impressive GUI. We support drag & drop, wysiwyg, internationalization (to some extent), office integration and we can taylor the system to our clients specifications easily. But compared to any windows/mac wordprocessor of the last 15 years it is actually pathetic in its abilities. Responsiveness of the UI is a problem (new screen means noticable delays while the browser is talking to the server, complex dom stuff also has some consequences). Basically most interactions require some form of serverinteraction. Thanks to some javascipt voodoo almost noone completely understands in our company we can actually prevent redrawing the screen every time. Getting things to scale properly in one browser is difficult, getting it to scale properly in IE and Mozilla is taking lots of our time (but yes we support Mozilla & Firefox). We have dropped safari support because it doesn't support everything we need. Our clients are amazed at what is possible with the technology while we have to live with its inadequacies.

This is the context of this proposal and this is why the proposal is totally inadequate. MS seems to have a better grasp of what is needed. Avalon/xaml is much more ambitious than this. If we continue to pretend we don't need it it will just be there someday without a credible alternative. And people will build their applications in it because their clients will want all the prettyness & usability (whether they need it is an interesting but irrelevant philosophical discussion). Basically the whole notion of applications running in a browser is obsolete. The browser is just a toolkit for running applications on top of. Mozilla despite its ambitions in this area, is still nothing more than a toolkit for creating browsers (and similar applications).

So KISS is not the solution. The first company that will produce a powerfull, easy to use toolkit for building web applications will get a major portion of the market. Right now it looks like MS has a four year headstart in creating such a solution. It won't be perfect just vastly better than what is out there today. It has realizaed two important things: 1) CSS/HTML is a dead end 2) Anything less than what client applications can do today is unacceptable for web applications. The future is multimedia userinterfaces that happen to be hosted on an applicationserver and require no clientside installation beyond the run-time engine.