4 ms·

The only reason AMP is faster than most webpages is because Google preloads and prerenders the content when you are still on the Google results page. It's not a

The only reason AMP is faster than most webpages is because Google preloads and prerenders the content when you are still on the Google results page. It's not a fair contest. We've seen our HTML pages be consistently faster than our AMP pages when all else is even.

jsnell 7y ago
But the whole point of AMP is to make prefetching possible in a safe and privacy-preserving way. The AMP caches are where the big wins come from, and what clearly guided the rest of the design. If option one makes prefetching possible and option two doesn't, why would a comparison with prefetching be unfair? (Signed exchanges could in theory change this and allow safe cached serving and prefetching of arbitrary HTML pages, but the crowd that hates AMP seems to despise signed exchanges even more. Such is life.) Incidentally, there was just a large scale study made on the performance characteristics of AMP in the real world. https://blog.apnic.net/2019/08/08/amp-up-your-mobile-web-experience/ https://blog.apnic.net/2019/08/08/amp-up-your-mobile-web-exp... TL;DR is that the AMP pages from the origin were loading on average in 3.1 seconds compared to 7.7 seconds for the matching HTML. (Not all sites are as diligent as you about keeping their HTML light). And then pre-rendering sped that 3.1 seconds to 1.2 seconds. Though the bandwidth used for the prefetching was more than I would have expected.
- auslander 7y ago
  > But the whole point of AMP is to make prefetching possible in a safe and privacy-preserving way. Scripts forced on us by biggest Ad company are privacy-preserving? Was it sarcasm?
  - jsnell 7y ago
    No. I could elaborate, but given the tone of your comment it seems pretty obvious you have no interest in that.
    - craftinator 7y ago
      I think that's the same tone of most comments on these threads. I don't think the HN crowd believes Google will do anything positive for internet users as a whole, unless it's also positive for Google's bank accounts. Hard to be a knight in shining armor when you've proven yourself a scoundrel, ya ken?

jsnell 7y ago

But the whole point of AMP is to make prefetching possible in a safe and privacy-preserving way. The AMP caches are where the big wins come from, and what clearly guided the rest of the design. If option one makes prefetching possible and option two doesn't, why would a comparison with prefetching be unfair? (Signed exchanges could in theory change this and allow safe cached serving and prefetching of arbitrary HTML pages, but the crowd that hates AMP seems to despise signed exchanges even more. Such is life.) Incidentally, there was just a large scale study made on the performance characteristics of AMP in the real world. https://blog.apnic.net/2019/08/08/amp-up-your-mobile-web-experience/ https://blog.apnic.net/2019/08/08/amp-up-your-mobile-web-exp... TL;DR is that the AMP pages from the origin were loading on average in 3.1 seconds compared to 7.7 seconds for the matching HTML. (Not all sites are as diligent as you about keeping their HTML light). And then pre-rendering sped that 3.1 seconds to 1.2 seconds. Though the bandwidth used for the prefetching was more than I would have expected.

auslander 7y ago
> But the whole point of AMP is to make prefetching possible in a safe and privacy-preserving way. Scripts forced on us by biggest Ad company are privacy-preserving? Was it sarcasm?
- jsnell 7y ago
  No. I could elaborate, but given the tone of your comment it seems pretty obvious you have no interest in that.
  - craftinator 7y ago
    I think that's the same tone of most comments on these threads. I don't think the HN crowd believes Google will do anything positive for internet users as a whole, unless it's also positive for Google's bank accounts. Hard to be a knight in shining armor when you've proven yourself a scoundrel, ya ken?