The Billion Dollar HTML Tag | Hacker News Reader

The Billion Dollar HTML Tag

datacenterknowledge.com

61 points1SockChuck29 comments

I always find this peculiar when google says they found users were doing fewer searches when they increased the # of results on the serp to 20 or 30. Maybe when a user sees more results they find what they were looking for more quickly, and so don't have to do another search. Yet google always turns this around and says this is bad because users do fewer searches. I guess from google's perspective, revenue-wise, it's bad. But from a user's perspective, it's good. Do no evil?

I don't think that hitting the 'next page of results' counts as a search. And whether they display 10 results initially forcing the user to hit 'next' twice to see all 30 results instead of just displaying 30 results outright... the search result set is still the same. If those 30 results don't contain what the user is looking for, they will still have to perform another search.

I guess this is assuming that people don't just re-search with different terms if what they are looking for isn't right there on the first page.

> I guess this is assuming that people don't just re-search with different terms if what they are looking for isn't right there on the first page.

Which I do all the time. I don't dispute that faster=better, but I have always been skeptical of Google's claim of just how much better.

Whenever Mayer trots out this same story, I think the same thing: "Why of course they search less, they find a good-enough result past #10, before either reformulating or next-paging!"

That makes a lot more sense to me than her usual spin, which is that the extra lag discourages more searches.

There's are reasons Google might prefer reformulation and next-paging beyond just "ad impressions": it gives better feedback about user's intent for training relevance algorithms.

That is, they learn more from your second query formulation to get a better result than if you just find a good-enough result at #11.

Or, if you open multiple results in background tabs before deciding the best, by forcing you to express click-intent to get past #10, they get a stronger signal as to whether the top ten results were good enough.

(I always set my result count to 100, and often open 3+ results based on snippets, some from below #10, before viewing any of the target pages.)

"Mayer consulted the W3C HTML specs and found a tag (the align=right table attribute) that would allow the right-hand table to load before the search results, adding a revenue stream that has been critical to Google’s financial success."

Is there any way to independently confirm that she was the one who consulted the spec and made this discovery? Because otherwise it sounds an awful lot like an executive taking credit for the work of someone underneath them.

It could also be a journalist sloppily attributing the work of someone underneath her. Also, though, she's pretty technically deep; she was doing AI before she started doing usability at the big G, and it wouldn't be at all surprising for her to come up with an idea like that, especially given that we're talking about a time when Google was a small company (this was when they were first adding AdWords; would have been in 2000 or 2001?)

Under some circumstances you might argue that there's more leverage in an executive like Marissa delegating work like this to someone else. But maybe sometimes the raw brainpower you bring to bear on a technical problem can really make a billion-dollar difference.

They actually found a way to draw the Google checkout cart using HTML - how would they do that?

The solution this article references was basically a table with 1x1 cells that had their background colors set via CSS classes. I was on the team with the guy who did the hack when I was an intern. That was about three years ago.

At some between then and now, Google actually went back to using a real image for it, and now it looks like it's just a CSS sprite.

The Google logo on the results page is part of a larger image. They use background-position to offset the part that you actually see.

http://www.google.ca/images/nav_logo4.png

For example, the value for the next arrow at the bottom of the page is: "background-position: -76px 0pt;"

Some of the buttons on there don't look too great.

I guess it is easier to notice when they isolated.

They are drawing the image in CSS as a background image to a span tag.

The tag: <span class="cps ss"></span>

The CSS: .ss { background-color: transparent; background-image: url(/images/nav_logo4.png); background-repeat: no-repeat; background-attachment: scroll; background-position: 0pt 0pt; height: 26px; display: block; }

Probably 1px row/columns/divs.

Does anyone know where to find an example? (of the Google Checkout logo)

Just do a search for something you'd like to buy like: http://www.google.com/search?q=computer and typically one of the sponsored links will have the icon.

/me waits for PB to tell us how it really happened.

A similar example is how Google renders the histograms on the Google Stock Screener - they use tables instead of an image.

http://www.google.com/finance/stockscreener

I don't believe Google would have been out a billion dollars if the ads had been on the left instead of the right.

Tell that to any AdSense publisher who has increased revenue 20-30% by moving ads a few pixels left or right, or by swapping the link color from one shade of blue to another. There are plenty. And Google is the ultimate AdSense publisher :)

There are plenty? Please cite sources. 20-30% improvements sound too good to be true for such minor alterations.

It's pretty much universal with AdSense. If you know anyone who makes a significant amount of money, ask them about it. Or just browse the AdSense board on WebmasterWorld.

I have personally increased revenue by over 100% by making small color and positioning changes, and this is also fairly common (I cited 20-30% because I wanted to be super-conservative in my claim).

But here are some examples:

From Google's official blog: Removing a border from an ad doubled revenue to $650/day:

http://adsense.blogspot.com/2006/05/unobtrusive-ads-can-boos...

Palette swap doubled clickthrough: http://www.cyberindian.com/web-marketing/article.php?article...

Webmasterworld thread: Most people did not disclose percentage, but some did (85% increase, 100% increase, etc)..

http://www.webmasterworld.com/forum89/14556.htm

There are endless examples.. just Google it.

The HotOrNot talk at CodeCon cited improvements of much more than that for a slightly more drastic change: moving the banner image to between the person's face and the rating bar, instead of above the rating bar. Unfortunately I don't remember the amount of the difference.

The billion dollar tag should be <resources>

A way to include external resources in html (images, audio, video, etc) in order to make a single file, which can be downloaded as a single file (same as MHTML, MAFF, etc but without the hacks)

I hope it gets considered and included in HTML6

Implementation:

- place it after the html closing tag to unclutter the file:

- use it from any html tag <img src="url(#myimg)"/>

  </html>
  <resources>
    <resource id="myimg" content-type="image/jpeg" encoding="base64">...</resource>
    <resource id="x0" type="stylesheet" content-type="text/css">...</resource>
    <resource id="x1" type="script">...</resource>
    <resource id="x2" type="font">...</resource>
    <resource id="x3" type="audio">...</resource>
    <resource id="x4" type="video">...</resource>
    <resource id="x5" type="object">...</resource>
    <resource id="x6" type="embed">...</resource>
  </resources>

That's a horrible idea and totally destroys the scalability of the web because it destroys effective caching.

That every resource has it's own url and is downloaded separately is not a mistake that needs corrected, it's what makes the web scale and it was done that way on purpose.

I am talking about "Save As..." one html file.

Take a look at your browser's File menu, and save this page as one single html file.

Open that file in any editor and take a look inside it.

Do the same for every browser out there, and then, only then, you'll understand my point.

Ah, you weren't very clear then, that's not at all what it looks like you're saying.

What makes that better than MHTML, which is basically a standard multipart/related MIME document as used by bazillions of email messages a day?

Simplicity.

It can be parsed and traversed like a DOM object.

It can be manipulated with javascript.

Unlike MHTML...

That's equivalent to data: URLs if you only refer to the object once in the HTML.

More or less, with the convenience of reuse, and a better location at the end of the document for better readability.

A more interesting tag would be <definitions> where you can define your own tags for internal use:

    :
    <menu />
    <businesscard />
  </html>
  <definitions>
    <define tag="menu">my menu here</define>
    <define tag="businesscard">my business card here</define>
  </definitions>

Most browsers nowadays allow any kind of tags, ignoring them, so all you have to do is set their style to hidden for unsupported browsers.

Now the semantic web takes a new meaning on its own.

Reply on news.ycombinator.com