Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Reactive prefetch on Google Search: 100-150ms speedup

plus.google.com

Post reply on HN
Print view View original

81–90 of 93 posts

Re: Reactive prefetch on Google Search: 100-150ms speedup

#81

Post by nitrogen »

logicallee wrote: post #77
nly wrote: post #44

Earlier quoted context omitted.

So when I google 'suicide' it's OK for Google to know whether I'm clicking on the Samaritans or the Wikipedia article? And when I Google 'rape' it's OK for Google to know whether I click on a news article about a string of recent rapes or click through to rape fetish erotica website? Not everything is black or white.

To your first question: YES, it's 100% okay for them to receive this query (a distinct issue from 'Google knowing'). I think in this case it's pretty black or white, because people would ordinarily take the time to add to their search queries. If people want to see results on (whatever fetish), do you think they would just Google the word 'sex' and then go from there, under cover that they might have just been lookin…

it's like the server in a restaurant asking you if you enjoyed your meal...

...and then being able to get you fired, steal your identity, or otherwise affect your life without your knowledge because you happen to enjoy a certain kind of food your boss doesn't like.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#82

Post by logicallee »

nitrogen wrote: post #79
logicallee wrote: post #76

Earlier quoted context omitted.

wow, so, this is what in my mind I was caricaturing. Offline living is basically incompatible with the knowledge available on the Internet, and what you mention are super unreasonable steps for normal users to take.

super unreasonable steps for normal users to take. Only because people have "forgotten" how to build offline apps.

no, it really is super unreasonable.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#83

Post by Illniyar »

Other then search engines what type of sites gain any benefit from this?

For external links, if you aren't a search engine, you don't know what resources need to be prefetched and even if you did you will have no way of knowing when it changes (other then manually). For internal links, in most cases the resources are already cached and http 2.0 server push is going to fix the rest.

Beyond that, the API looks clunky - instead of having a "prefetch" attribute on a link, you need to add more links with special syntax to the header dynamically.

What's the point in adding this functionality to a browser? this looks like an implementation specifically designed to make only google search faster.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#84

Post by logicallee »

nitrogen wrote: post #81
logicallee wrote: post #77

Earlier quoted context omitted.

To your first question: YES, it's 100% okay for them to receive this query (a distinct issue from 'Google knowing'). I think in this case it's pretty black or white, because people would ordinarily take the time to add to their search queries. If people want to see results on (whatever fetish), do you think they would just Google the word 'sex' and then go from there, under cover that they might have just been lookin…

it's like the server in a restaurant asking you if you enjoyed your meal... ...and then being able to get you fired, steal your identity, or otherwise affect your life without your knowledge because you happen to enjoy a certain kind of food your boss doesn't like.

Yes, exactly. And not even 'a certain kind of food' but rather which food that you had already ordered.

The metaphor is spot-on. "Did you enjoy your meal"? "Oh so you can get me fired, steal my identity, or otherwise affect my life without my knowledge because I happen to enjoy a certain kind of food my boss doesn't like? No comment."

The reason it's a good analogy is because you already 'ordered' (the search terms are the meaningful data) and a statistical sampling of which of the top 10 results for "octopus hentai videos" a random sampling of users who entered that search term clicked on, is not in practice used for anything other than improving ranking quality.

it's exactly the same as "did you enjoy your meal" or what you liked about it - you've already given up most of the info by ordering in the first place.

so your analogy is a good one. it's just completely innocent.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#85

Post by ce4 »

logicallee wrote: post #82
nitrogen wrote: post #79

Earlier quoted context omitted.

super unreasonable steps for normal users to take. Only because people have "forgotten" how to build offline apps.

no, it really is super unreasonable.

I disagree. I'm not a luddite. First, I didn't say that I use that exclusively. And furthermore:

* it's nice to have knowledge at hand when offline (commuting via train, abroad w/o a local data plan, no gsm coverage, edge connection instead of 4G)

* saves bandwidth for my 1GB/month data plan when appropriate

* Aard dict can be used to evade filtering (not my primary concern, but there are people in other countries who might benefit)

Re: Reactive prefetch on Google Search: 100-150ms speedup

#86

Post by malthaus »

chii wrote: post #72
crazy1van wrote: post #50

Earlier quoted context omitted.

I am all for privacy and am a daily user of encrypted email and encrypted chat. However, I think there are some activities that the very privacy minded need to accept that they simply cannot participate in and maintain complete privacy. Search is one of these. In order for search to work, the user has to send a query to a third party. The third party has to be able to read the query to do the search. If that's not an…

the argument here is that the search engine should not be privy to which result you selected.

so how are they supposed to improve/tune search results? that's the first metric i'd like to know.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#87

Post by cmdkeen »

chii wrote: post #72
crazy1van wrote: post #50

Earlier quoted context omitted.

I am all for privacy and am a daily user of encrypted email and encrypted chat. However, I think there are some activities that the very privacy minded need to accept that they simply cannot participate in and maintain complete privacy. Search is one of these. In order for search to work, the user has to send a query to a third party. The third party has to be able to read the query to do the search. If that's not an…

the argument here is that the search engine should not be privy to which result you selected.

Then use a search engine which offers that functionality. You'll probably find the results are worse than Google.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#88

Post by walterbell »

logicallee wrote: post #82
nitrogen wrote: post #79

Earlier quoted context omitted.

super unreasonable steps for normal users to take. Only because people have "forgotten" how to build offline apps.

no, it really is super unreasonable.

Travelers with intermittent connectivty are best served by offline databases with async/push updates.

Re: Reactive prefetch on Google Search: 100-150ms speedup

#89

Post by 3minus1 »

Illniyar wrote: post #83

Other then search engines what type of sites gain any benefit from this? For external links, if you aren't a search engine, you don't know what resources need to be prefetched and even if you did you will have no way of knowing when it changes (other then manually). For internal links, in most cases the resources are already cached and http 2.0 server push is going to fix the rest. Beyond that, the API looks clunky -…

> you don't know what resources need to be prefetched and even if you did you will have no way of knowing when it changes

you could find out if google set up an API that exposed this information. Then a simple plugin could update the links on your page.

Post reply on HN