JavaScript Features To Forget

12 May 2020

The first demo of the language that was to become JavaScript took place almost exactly 25 years ago.

The language was released, as LiveScirpt, in a beta of Netscape Navigator in the fall of 1995, and renamed to JavaScript later that year. Sometime late that year I began work on the first edition (O’Reilly published it as the “beta edition”) of JavaScript: The Definitive Guide, and it was published in August of 1996, meaning that it is going on 24 years old now.

With the seventh edition coming out in just a few weeks, I want to take a trip down memory lane and blog about some old weird features of JavaScript and of the early web platform that we can now, mercifully, forget about. One of the reasons that the new 7th edition is thinner than the 6th edition is that I’ve removed the reference section. But in addition, I found that there was a lot of stuff that was simply no longer relevant to web developers in 2020. Web compatibility is forever (or for at least more than 25 years) so browser vendors may need to keep supporting old and obscure language and platform features for a long time to come. But there is no need for the rest of us to clutter our minds with them anymore.

So here are some JavaScript and Web platform features that no longer bulk up the page count of JavaScript: The Definitive Guide. I’m happy to say goodbye to them. (I feel myself winding up for a bit of a rant here, so be warned that this is not carefully researched; just my memories of how things were in the bad old days):

  • The arguments object has been completely obviated by the introduction of ...args in ES6. It was always difficult to explain the weird way it interacted with named arguments and to always caution about its performance implications. You may still see it in legacy code and you may be reminded of its presence if you try to name a local variable or function parameter arguments in strict mode, but now that we have rest arguments, it should be allowed to slink into oblivion.

  • We used to have to worry about the performance of repeated string concatenation. There was a period of where we all learned to push strings onto an array, and then use join() to concatenate everything at the end. Then JavaScript got fast and we were all “Well, actually” and unlearned that pattern. And now with template literals, who even uses string concatenation anymore!

  • document.write() was pretty much the main feature of JavaScript at the very beginning, before the DOM. (If you did not use JavaScript in the 20th century, you may not know that there was a time before the DOM, but it is true. There was a property you could set to change the background color of a web page, but no way to actually alter the document tree once the document had been parsed.) IIRC you could even insert scripts into a document with document.write(), but you had to be careful to break the closing </script> tag into two strings so that the HTML parser wouldn’t interpret it as the end of the currently-running script.

  • HTML didn’t have an <iframe> in the early days, but there was <frameset> and <frame>. The window.frames property was an array of the nested window objects that represented the frames in a document. It was possible to actually call the open() method of a document in a frame, and then use document.write() to dynamically generate an entire document inside that frame. That was kind of cool, actually. Because frames could be nested inside other frames, every Window object had a frames array that contained its children frames, and a parent property that referred to the containing window, and a top property that referred to the toplevel window. Earlier editions of my book dedicated long sections and complicated figures to explaining all of this.

  • There are all kinds of obsolete techniques for referring to particular elements within a document. The frames array was one, but, IIRC, there were also links and images arrays as well that were literally just a list of all the links and images in a document. IE (version 4, I think) went all-in and introduced document.all: an array of all elements in the document. (This was the beginning of the DOM and of “DHTML”—kind of like the first fish that crawled onto dry land.) document.all had all kinds of weird features: it was an array that also had methods for looking up elements by name or something like that. document.all was never standardized, but even the standard methods like document.getElementById(), document.getElementsByName(), document.getElementsByTagName() and document.getElementsByClassName() seem obsolete today, crushed into irrelevance by jQuery’s $() function and the standard document.querySelector() and document.querySelectorAll() methods that it inspired. Through the power of CSS selectors those two functions obsolete everything that came before.

  • The thing I hated most about Internet Explorer was that it used an attachEvent() method for registering event handlers. In my memory, they did this even though the standard addEventListener() had already been defined, and that really annoyed me. Events and event handling were one of the biggest sources of incompatibility on the web, and for years JavaScript programmers (and JavaScript book authors) had to deal with a lengthy list of differences between the IE event model and the standard event model. Event handling code had to be written twice: once for IE and once for Firefox. Book chapters about events were twice as long as they needed to be because there were two similar, but completely incompatible ways of dealing with them. One of the major features of jQuery was that it implemented its own event compatibility layer so you only had to know about jQuery events, and I suspect that this was an important reason for its popularity.

  • The original DOM API was defined during the era of magical thinking about XML. (People really seemed to believe for a couple of years that XML was going to solve all data problems. It was a strange time.) Somehow the W3C committed that defined the DOM API got infiltrated by Java people who felt it was appropriate to define a single API to be used by JavaScript programmers working with HTML documents and Java programmers working with XML data. That’s why we have weird things like Attr nodes (which are best ignored). One of the things that always bugged me about the DOM Level 3 API was that to remove an element e from the document, you couldn’t just write e.remove() as we would today. You actually had to write e.parentNode.removeChild(e).

Anyway, it is 2020 now and I promise that the new 7th edition of my book will not subject you to lengthy explanations of any of those old features that are best left forgotten.