Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

As someone who thinks HTTP2 is garbage and refuses to support it in my enterprise, this is a good step towards a MUCH better solution.


I don't know much about HTTP2; why do you dislike it?


one reason to dislike it, is because it is way too stateful. and a lot of things depends on timings. /and the spec misses some stuff, like if all my streams are done (except one long running stream, SSE, etc) and I get a new request what do do now? (there are multiple ways of doing that, one would be to just cancel the long running stream and hope that the client reconnects to the new connection).

Personally I think it is more complex than it was needed to be. stream priority can also be a PITA.

Ah and I basically forgot, due to the state it is practically not the best protocol for mobile clients over 2G/low end 3G.

It's not the worst, I think h2 would be fine if it would be 10 years earlier, since I think we can learn a lot from the behavior of h2 in real world scenarios and than use the best things from http/1.1 and h2 to make h3.


I thought HTTP2 was born out of the experimental SPDY protocol which was used in the wild for quite some time, maybe 5 years. I think SPDY was released around 2010.

HTTP2 is like the good parts of SPDY and what was learned over that time period (I think).


SPDY wasn't in widespread use (I think it had something like 30% usage), especially not in mobile environments. basically I also did not say that h2/spdy is bad, it actually solves some things, but makes others worse. it actually solves head of line blocking. multiplexing is cool, but comes with the overhead of a connection state.

I think h2 is perfect for RPC like architectures, interconnection between servers is a perfect fit for something like h2. but when it comes to real world traffic it depends on too many parameters. The sad part is, is that h2 will take forever to gain that data and to have widespread usage. the hard part will be 100% h2, a lot of stuff prolly only uses h2 at the edge.

Also most flaws of h2, will probably be fixed without a new protocol, i.e. in the next 10 years I hope that _all_ mobile environments have a widespread use of 3g networks (or to say it differently, I dream of) and hopefully all network people make world wide latency better and make all networks more reliable.


Just bear in mind the security considerations: https://tools.ietf.org/html/draft-ietf-httpbis-early-hints-0....

This is not particularly safe to use in an HTTP/1.1 environment unless you do UA sniffing.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: