3 ms·

Has anyone pulled this thing apart? I am curious how they can claim 5x speed improvement for JavaScript compared to Safari. AFAIK third party apps still cannot

by st3fan 13y ago

Has anyone pulled this thing apart? I am curious how they can claim 5x speed improvement for JavaScript compared to Safari. AFAIK third party apps still cannot use a JIT because they cannot execute dynamically generated code. Wonder what is behind those claims.

hnriot 13y ago
It's probably the same tech as the kindle cloud acceleration tech. Basically the server (cloud) accelerates the page by fetching all dependent objects in parallel on th faster network, then send the result over a single compressed channel. This saves the chatty nature of http and if modified since requests and of course the startup/tear down times per origin server.
zapt02 13y ago
It looks like a lot of stuff is prerendered. Check their pro version which offers "unlimited Adobe Flash minutes". I guess it's a VNC-like tunnel.
azakai 13y ago
They claim to have a "revolutionary JavaScript engine" that is 5.5x faster on sunspider than chrome on android and safari on ios. The only way that could possibly be true is if they run the JS on the server, which has a faster CPU than a mobile device.
- mistercow 13y ago
  If so, they're being extremely disingenuous. Sure, you might be able to run those benchmarks faster, but the latency would eat up all of your gains and then some in any practical scenario.
0xdripdrop 13y ago
Yes, we reverse engineered Puffin 2.x and implemented a partially featured client in python. We used this client to run map reduce on puffin's servers. You can read about some of the benchmarks we measured here: http://ix.cs.uoregon.edu/~butler/pubs/acsac12.pdf http://ix.cs.uoregon.edu/~butler/pubs/acsac12.pdf The short answer to your question is they run and render all the js/html/css in a server farm (rackspace when we benchmarked it) and send the client a rendered stream of the contents. The paper above gives the benchmarks we measured. Note: all the above is only relevant to Puffin 2.x. Im not sure how much has been redesigned for 3.x None of the above may be correct for current versions.
- kintamanimatt 13y ago
  That's one hell of a serious privacy problem!

by st3fan 13y ago

hnriot 13y ago

It's probably the same tech as the kindle cloud acceleration tech. Basically the server (cloud) accelerates the page by fetching all dependent objects in parallel on th faster network, then send the result over a single compressed channel. This saves the chatty nature of http and if modified since requests and of course the startup/tear down times per origin server.

zapt02 13y ago

It looks like a lot of stuff is prerendered. Check their pro version which offers "unlimited Adobe Flash minutes". I guess it's a VNC-like tunnel.

azakai 13y ago

They claim to have a "revolutionary JavaScript engine" that is 5.5x faster on sunspider than chrome on android and safari on ios. The only way that could possibly be true is if they run the JS on the server, which has a faster CPU than a mobile device.

mistercow 13y ago
If so, they're being extremely disingenuous. Sure, you might be able to run those benchmarks faster, but the latency would eat up all of your gains and then some in any practical scenario.

0xdripdrop 13y ago

Yes, we reverse engineered Puffin 2.x and implemented a partially featured client in python. We used this client to run map reduce on puffin's servers. You can read about some of the benchmarks we measured here: http://ix.cs.uoregon.edu/~butler/pubs/acsac12.pdf http://ix.cs.uoregon.edu/~butler/pubs/acsac12.pdf The short answer to your question is they run and render all the js/html/css in a server farm (rackspace when we benchmarked it) and send the client a rendered stream of the contents. The paper above gives the benchmarks we measured. Note: all the above is only relevant to Puffin 2.x. Im not sure how much has been redesigned for 3.x None of the above may be correct for current versions.

kintamanimatt 13y ago
That's one hell of a serious privacy problem!