News for hackers. Stuff that matters.
Re:Firefox is back. It's time to give it a try by skywhopper on Thursday June 21, 2018 @09:01AM (#17364654)
The power of a company like Google can be great when harnessed appropriately. When Chrome was the new kid on the block, they had every incentive to be compatible, faster, better, more flexible, etc. Now that they are the leader, their incentives are very different. I'm very concerned about the rapid move away from straightforward text-based protocols to opaque binary protocols. Yes the performance is better, but the protocol itself is far less flexible, harder to troubleshoot, and harder to build on. The whole reason HTTP took off was because it was simple, human-readable, and easy to implement, and built on TCP which was well understood and simple. It was a liberating change from the rigid, corporate, non-standard binary protocols coming from the Microsofts and IBMs and Oracles of the world. There are plenty of ways HTTP/1.1 could have been improved for better performance without sacrificing its fundamental openness. But HTTP/2 and QUIC are taking us back into the realm of non-standard and more rigid and opaque binary protocols. Performance on poorly designed sites that load tens of megabytes of crap from hundreds of URLs on each page load is the hook Google is using to push us back into what I consider a dark age of network communication. But it doesn't have to be this way.
Re:Firefox is back. It's time to give it a try by lallysingh on Thursday June 21, 2018 @10:18AM (#17365330)
HTTP/2 is standard and isn't opaque. You can use any number of high quality tools to dump it -- here's one: https://wiki.wireshark.org/HTTP2 .

The performance improvements are quite substantial.

Re:Firefox is back. It's time to give it a try by oscargrouch on Thursday June 21, 2018 @03:31PM (#17368520)
> But HTTP/2 and QUIC are taking us back into the realm of non-standard and more rigid and opaque binary protocols.

This whole BINARY vs. PLAINTEXT paradox where the binary tends to be more efficient where the plaintext more open and intermediate ground between humans and machines, for me can be solved in a optimal way where the final output is always binary, but where you can assist humans with plaintext latter.

Just look how many beautiful and expressive computer languages we have now where the output is a pure obscure binary.

So i think if you follow this path you can have the best of both worlds. And because of that i tend to disagree with your point of view, where the default must be plaintext.. The layered approach is probably much more sophisticated and less amateurish when you care about waste of CPU cycles and RAM memory, without the need to have only a obscure and opaque representation.