Here's my pet peeve that I don't see on the list: not giving me the results for what I searched for the first time around. For example, yesterday I searched for "OCUnit coverage" or something similar and Google said "showing results for 'JUnit coverage'". (I don't mind suggesting another search query, but I expect the first hit to assume that I didn't mistype.)
I actually like that feature. I generally tend to know how my search queries sound but not how they are spelled, so this gets me to where I want to go faster. I imagine this is also true for a large number of people, so I doubt it would be changed.
In fact, he does not put forth any argument at all. There is talk of the fact that it was indeed ruled 'fair use', and he essentially just disagrees and considers it evil, and expects that to have some weight.
I fundamentally disagree: it is fair use, I have personally found the feature useful more than once, and I really just don't see it as evil. As he mentions, publishers extra-concerned about their content and copyright can opt out; for everyone else, it has let the web be a little more stateful. Letting people disappear controversial content from pages early on in a public backlash just does not strike me as having any benefit.
I don't believe it is fair use to make an entire copy of someone else's material, if you've made no transformative change to it.
I understand there is an opt-out. My post on daggle.com goes into great depth about that. However, I don't think that's the way Google should operate.
Google just assumed that it was OK to reprint material through its cached pages; many have felt that was presumptuous. So far, they've been OK with it -- but it remains something that colors them as evil in some quarters.
I get you don't agree with that, but others do view it that way. And it sure jumped out to bite Google when it levied allegations that Bing was copying Google.
>Google still does plenty of testing. And we still get messages from people who wonder if they’ve been hit by malware. But it feels like this is happening less. Ideally, the company would regularly advise people of when a test is underway.
Have they not heard of "New Coke"? Constant, minor flux is accepted much more readily than periods of non-change followed by a big change.
The author suggests charging a token amount (eg $1) in order to discourage junk. This would certainly have discouraged me from using my (Blogger) blog to send pictures and news home to family and friends when I go travelling. Not that paying a nominal fee would be a problem in itself. Rather, the administrative overhead of making the payment would have driven me to Wordpress or something else. The cost of making a small payment is not just the value of the payment, it is the cost in terms of time of making the payment and the time spent checking my credit card statement. I certainly can't agree with this suggestion, especially when Blogger has so much competition.
Point #1 annoys me a lot. It would be possible to perform many interesting analyses on your own (think Venn diagrams, association graphs on natural language) if the numbers were accurate. But they are not.
I actually heard first-hand from a Google employee an explanation for why these numbers are like that. She said something like, "they are only estimates." Well the point is, if these estimates are as inaccurate as they are, why put them up at all? Doing so is almost dishonest-like.
I almost have a feeling (warning: getting into conspiracy territory here) that the search recall numbers are inaccurate on purpose, so that no third-party analytics could be built using them.
I tend to agree. SEO is a disease that makes Google lots of money, but is destroying the ability to find anything. Keywords simply do not work anymore. I have a feeling that there are probably 5-10 startups out there working on this and at least a few are going to come up with some seriously compelling alternatives. The search+ads sector is absolutely RIPE for disruption.
This was a good post. #2 was the most annoying to me. I noticed that the main problem has to do with Chrome remembering your user behavior which affects your results.
Most of these flaws are trade-offs, where if you fixed it, you'd create a worse problem.
This is most obvious in the ones where Danny's judgment itself may not match many of the readers of the article. For example, I love cached pages (as a user. As a Google dev, I kinda hate them). They've been so useful in cases where a website has unexpectedly gone off-line or becomes overloaded. But Danny's a publisher, and as such, each cache hit is one less pageview that his website gets. His interests are opposed to those of users, and Google, as the party in the middle, has to arbitrate between them. There's no right choice that will avoid pissing off one or the other.
Or take one of his LOVEs: "An end to overwhelmed product launches." A direct result of that has been that it's become much harder to launch things, and there is correspondingly less innovation coming out of the Googleplex. Every service needs to build for scale from day one, and that means that you have to pay that tax and iterate more slowly even if your service won't get any users. If a startup tried that, they'd never get off the ground.
Assumptions. Engineers are confident that building a search engine requires thousands of servers, indexing the entire web, coming up with some super complex algorithm etc...
Maybe we need a couple of guys who do not know their limitations and are ready to approach search from a different angle.
hate: google places (silly aggregator site hArdcoded into the google serps)
hate: google custom seatch api (with a hundred (!!!!) searches free / day) instead of the proper google search api