4 ms·

AMP preloads pages before you click them, which means it's much faster than mobile. You could argue that a browser could preload Google search results before y

AMP preloads pages before you click them, which means it's much faster than mobile. You could argue that a browser could preload Google search results before you click them, but: (1) those pages could be giant, and loads 10 giant pages would be a bad bandwidth tradeoff and (2) there are privacy implications, since this would leak information about what users are searching for, even if they don't click the results.

lucb1e 9y ago
> and (2) there are privacy implications, since this would leak information about what users are searching for, even if they don't click the results. This is wrong. Google can proxy the traffic like they do with Gmail, negating that privacy issue. I was typing a comment in which I said that I'd prefer your privacy risk (where I leak info to random sites that I might want to visit anyway) over the current privacy issue (all sites using AMP leaking data about me to Google), but it's worse than that: the risk you identify is easily mitigated by proxying the content. The other risk you mention is not as easily mitigated. Proxying might help in part, because then the servers can be selective and/or stop at a certain point. Another idea is to only preload the HTML, and perhaps a limited number of resources smaller than a certain size (mainly css and small-size images, since JS is usually not used in styling, so preloading html and css should give a first, static impression before the rest can be loaded). All of this would have been better than any part of what they chose. I can see only two reasons for Google not to choose this: (1) more control over the web (this I think is true) and (2) more tracking (this I think a few people within Google might see as a nice side-effect, even if the vast majority of employees has only good intentions, so I don't think it was a driving reason but it probably contributed).
- ec109685 9y ago
  Google can't serve anything that they proxy on a google domain, so they need something like cdn.ampproject.org to cache the content. They also need to guarantee the content will behave properly when cached, hance the AMP format.

lucb1e 9y ago

> and (2) there are privacy implications, since this would leak information about what users are searching for, even if they don't click the results. This is wrong. Google can proxy the traffic like they do with Gmail, negating that privacy issue. I was typing a comment in which I said that I'd prefer your privacy risk (where I leak info to random sites that I might want to visit anyway) over the current privacy issue (all sites using AMP leaking data about me to Google), but it's worse than that: the risk you identify is easily mitigated by proxying the content. The other risk you mention is not as easily mitigated. Proxying might help in part, because then the servers can be selective and/or stop at a certain point. Another idea is to only preload the HTML, and perhaps a limited number of resources smaller than a certain size (mainly css and small-size images, since JS is usually not used in styling, so preloading html and css should give a first, static impression before the rest can be loaded). All of this would have been better than any part of what they chose. I can see only two reasons for Google not to choose this: (1) more control over the web (this I think is true) and (2) more tracking (this I think a few people within Google might see as a nice side-effect, even if the vast majority of employees has only good intentions, so I don't think it was a driving reason but it probably contributed).

ec109685 9y ago
Google can't serve anything that they proxy on a google domain, so they need something like cdn.ampproject.org to cache the content. They also need to guarantee the content will behave properly when cached, hance the AMP format.