Really cool reading this back story, in 2010 I launched a web search engine that displayed search results as favicons (no text, image only). The goal was to appeal to kids (using images and not text) and mobile devices (I generally find only 3 search results fit on a mobile screen using standard search engines, but the favicon engine allowed a user to see 30 search results on the screen at once)...the traction did not last and it was shut down. Nevertheless, great learning experience and my fondness of favicons and their importance only grew.
Edit: The part about Yahoo also made me recall a prior version on Yahoo's mobile search which displayed favicons prominently to the left of the search result text (first thing your eyes saw was the favicon), and at the time I felt Yahoo had the most aesthetically pleasing results page of all the search engines on mobile, all because the addition of the favicons.
yeah, I have no idea what our bookmarks bars would look like without favicons, especially if you used firefox's extension that shows just the favicon (which was one of the reasons I held onto firefox for so long).
There's a video somewhere of Mark Zuckerberg and some co-workers discussing ideas about Facebook (in it's earliest days), and in the video they are trying to decide on what options should be available under the relationship field, one guy says something like: "I feel like we should make it possible to say that.. like you're not in a relationship per se, but nor are you free, like it's complicated..."
This kind of ad-hoc innovation is something which is basically dead now.
Back then the number of people involved in browser and web technology was small. Now, even the brightest mind will have any new idea critiqued and iterated by thousands of equally bright minds whether they like it or not.
It's probably too much to say that there are no new ideas in web browsers any more, but it's certainly harder to find them thanks to how collaborative the Internet has made the world (that's almost ironic)
> noticed an unusual spike in HTTP requests for /favicon.ico
I remember a guy at Uni reacting to this in a fit of anti-MS pique ("How DARE they go making useless non-standard requests to my server behind my viewers' backs!!!!!!!") and putting a large meaningless file there for IE to find. It didn't take him long to work out he was only hurting himself...
This has some advantages over just blindly loading one file (no HTTP request if there's no image, though browsers still all try /favicon.ico nowadays) and you get to chose where to place the icon (probably together with other images of your site).
I think by now all browsers support <link rel="shortcut icon"> but all of them still request /favicon.ico so by adding that you a) don't have to change your markup for the icon and b) you silence the error log of your web server - all in one go.
> This has some advantages over just blindly loading one file (no HTTP request if there's no image, though browsers still all try /favicon.ico nowadays) and you get to chose where to place the icon (probably together with other images of your site).
Also, browsers which aren't MSIE can use PNG, GIF (including animated) or JPEG favicons.
> I think by now all browsers support <link rel="shortcut icon">
Only rel=icon is necessary in most browser, rel="shortcut icon" is for MSIE lte 9.
ICO's a perfectly reasonable format for small icons, and PNG was still a new format then.
I've written stuff that handled the ICO format; it's simpler and easier to load than PNG/GIF/JPG (unless you already have libraries for the latter, of course - which wasn't a given in browsers at the time). I don't think their choice was wrong at all, especially since ICO loading was at the time built into the OS, so any Windows browser that wanted to load favicons didn't even have to build their own loader.
EDIT: Plus, if you're complaining about 'favicon.ico', it's no worse than robots.txt. At the time the modern 'link rel="icon"' approach was definitely not in common use so it's no wonder nobody at MS chose it, and people probably would have complained loudly about MS introducing 'some proprietary icon HTML'.
Another reason ICO is good as a favicon format is that it makes it impossible for site authors to accidentally use a 500KB image file as their icon (which you could trivially do with png, gif, jpg, etc.) - ICO is intentionally a file format for small images, tailor-made for what icons were like in that era. So by relying on ICO they ensured that the favicon.ico on each site would be small, cheap to download, easy to cache, and easy to author.
And ICO has the huge advantage that you can provide multiple bitmap sizes (and color depths) in a single file, avoiding the bullshit of having to specify 6 different icons for 6 different size specs.
Technically, ICO is really a container more than an image format, the original ICO just contained bitmap data, since Vista you can also put PNG data in your ICO file (which MSIE recommends for 256x256 icons).
A similar alternative being Apple's (somewhat more flexible) ICNS
Edit: The part about Yahoo also made me recall a prior version on Yahoo's mobile search which displayed favicons prominently to the left of the search result text (first thing your eyes saw was the favicon), and at the time I felt Yahoo had the most aesthetically pleasing results page of all the search engines on mobile, all because the addition of the favicons.