Browser Bloat

It's that time of the year again. Or should I say "month?" Netscape has just released another upgrade to the popular Netscape Navigator Web browser. This upgrade, code-named "Atlas," adds a raft of new features to the familiar Version 2.0 browser, which was released in final form only a few weeks ago.

Atlas is the precursor to the next step in Navigator's evolution, Version 3.0. It's "alpha" code; in other words, that celebrated tower in Pisa is more stable. When it works, Atlas promises to deliver:

VRML viewing. VRML (Virtual Reality Modeling Language) is one of the competing standards for expressing three-dimensional information in compact form. VRML documents can be static or interactive. For example, you could create a Web page that hosts a virtual "village." Visitors could "explore" the village by clicking and dragging. Shades of Apple's e-World!
Audio playback. Prior to Atlas, to hear an audio file you had to install an external "audio player" like RealAudio, or save the sound file and play it with the Media Player. Navigator Version 3 will include a player for the most popular sound formats: AIFF, AU, MIDI, WAV. The new player is based on the Navigator Plug-In API, so it works seamlessly with the browser.
Atlas is also (supposedly) able to play back AVI motion video files, though I haven't been able to get this to work yet.
Phone. The alpha browser contains a very impressive add-on called CoolTalk. Eventually, CoolTalk will be able to provide live voice connections over the Net, much as Vocaltec's commercial "Internet Phone" does now. It's a little unstable at the moment, but the theory's there.
Chat. CoolTalk's "Chat" mode is a precursor of Internet telephony. Since the very earliest days of Unix, computers running that arcane operating system have included a "talk" or "chat" utility that allows two users to hold a real-time keyboarded conversation. What one user typed would appear on the other's terminal and vice-versa.
CoolTalk provides a windowed version of this venerable technology. Using it, you can converse--one keystroke at a time--with anyone else in the world that has access to an Internet connection and a Chat utility.
Whiteboard. This is one of the most fascinating innovations in the new browser: CoolTalk's "Shared Whiteboard" mode. Whiteboard is really fun to play with; a group of people can use simple tools to scrawl on a virtual drawing board; each person connected to the drawing "group" receives all the changes made by members of the group.

There are a number of other, less important improvements, such as: more logical "forward and back" navigation in frames; a new "background color" attribute for HTML table cells (what's another unilateral HTML extension between friends?); enhancements to the supplied USENET news and Net Mail clients; and some changes in the way Navigator interprets Web pages (hopefully intended to make the browser crash less often).

Bigger and bigger...

Taken one by one, each of these improvements to Navigator is valuable. I was especially impressed by Whiteboard: though developed by InSoft, it's the first thing I've seen from Netscape that wasn't based at least partly on ideas stolen from other companies. Compare Netscape's Usenet news browser to Forté Agent (or CoolTalk's Phone mode to Vocaltec's Internet Phone) and you'll see what I mean.

I'm not scolding Netscape for a little borrowing here and there; stealing good ideas is nothing new in the computer business--just ask Bill Gates. What I am asking, though, is, "Where does it end?"

The original Netscape browser download file (all those months ago) hovered at a bit more than a megabyte in size. It loaded quite quickly and required only a relatively small amount of memory.

The download file for Atlas is--all together now--six megabytes long. Assuming 100 percent throughput at 28.8kbps, something nobody has ever achieved, it would take 34 minutes to download. Given an averagely sluggish day on the Net, and factoring in Netscape's generally overloaded ftp sites, 45 minutes to an hour isn't unrealistic.

That's a long time to wait, especially considering the number of times you'll probably have to download the thing before a truly stable version emerges. Even better, based on recent history, as soon as Version 3.0 comes out, Netscape will put up an alpha of its next browser. Anybody think that will be smaller than six megabytes?

I suppose it's the sign of a maturing industry, though this one seems to have matured awfully quickly. In the early days of computing, everyone used one company's spreadsheet, a second company's word processor, and a third company's drawing program. Then Microsoft invented the office suite and everybody started buying all their software from Redmond.

We are seeing the beginnings of a similar trend in Internet products. I use Qualcomm's Eudora Pro for my Net mail, (mostly) Netscape for Web browsing, and Forté Agent for Usenet. Microsoft is promising to build all the Net functionality I need right into the operating system, and Netscape is racing to cover my every Internet need in one package.

The Downside

I admire the theory and deplore the practice. Netscape is adding features to Navigator at a frenetic pace, and we of the hoi polloi are paying the price. Each new feature makes Navigator bigger and destabilizes the existing features.

Netscape's motivation is twofold. First, they want to make their browser worth money. They have to; Microsoft is giving away Internet Explorer. The best way to do this is to build in features Microsoft doesn't have.

Second, Netscape's managers are taking a page from Microsoft's book. They hope that by being first, they can set and control the Internet's standards. While it's true that pioneers can blaze a trail, they also tend to find all the pitfalls. Sometimes it's better to be second and learn from the mistakes of those who went before, another strategy Microsoft has used to great effect.

The problem is that Netscape's two agendas can have conflicting, even opposing, objectives. To be better than Microsoft, Netscape needs a stable browser. Nothing's more frustrating than continual crashes. To be first, so as to set standards, Netscape has to ship software as quickly as possible, bugs or not.

The result is a browser that always has more potential than value. Netscape has had some great ideas, but their execution is erratic at best. Every release shows signs of being hurried. Simple little things, such as setting the display font size in one version, don't work correctly.

Diminishing Returns

There's another consequence of all this upgrading: each new feature serves a smaller population. For example, consider Java. A few months ago, Netscape doubled the size of the Navigator download file in order to add support for the Java programming language.

The trade press went wild. Everyone predicted that Java applets would soon be omnipresent on the Net, appearing as quickly as Web pages once did. Some of the industry's giddier pundits predicted that Java would soon eclipse C++ as the computer world's primary professional development tool. A few of the real fringe-dwellers even predicted that Java would cause the end of Microsoft's dominance of the desktop market. Of course, it all came true, just as when Jerry Pournelle predicted that the whole world would be developing with Modula-2 by 1988.

Actually, nothing changed. The few Java applets out there mostly fall into the "fun to see once" category. Netscape--and the pundits--fundamentally misread the market.

The reason the Web exploded is that practically anybody can develop a Web page. HTML had just the right combination of power and simplicity. In fact, HTML 2 (plus the "original" Netscape extensions) is still the standard most people develop their Web pages to.

Java's different. It's a programming language (actually a loopy dialect of C++), and it's hard. Compared to the enormous population of HTML developers, only a tiny handful of people have the knowledge and patience to develop Java applets.

Netscape made Navigator much bigger and more complex to serve what has so far turned out to be a very tiny need. With this size and complexity have come higher demands on system resources and more potential for instability.

And it didn't stop there. Netscape has added built-in support for VRML, at least partly to try to make that language the standard for 3D interactivity on the Net. As with Java, developing worthwhile VRML sites will take a lot more work than building simple HTML pages. Only people who have too much time on their hands (or those backed by healthy budgets) are likely to use VRML for some time to come.

Where will it end?

If we keep up at this pace, it'll soon take hours to download each Netscape Navigator upgrade, and the upgrades will start coming days (instead of weeks) apart. Nobody will know what features their current version has, and everybody will have a different set of bugs. The browser will require more memory than the operating system it runs on. The $500 "Internet Appliance" that hardware vendors keep promising will turn out to be an $8000 Sun workstation.

Enough already! Netscape needs to decide what features are really necessary and offer the rest as plug-ins and add-ons. To preserve user sanity, the base feature set should be frozen permanently. Anything beyond this feature set should be optional. Perhaps Netscape could even design a system to dynamically download extensions on demand.

Have your say!

What do you think? Click the button at the bottom of this article to respond. Once some comments have built up, I'll add a "reader replies" section to hold your opinions.

[Home] [Table of Contents] [Next] [Feedback]