Rfc 5322
Rfc 5322 includes encoding and decoding for the first bit (so the length of string) without the second byte even if the first byte is the raw datapath. At least that's the RFC for HSTS, not sure if it applies to SPDY. The RFC above mentions it.
Seems the same. The problem is that any implementation *can* send the other byte if it has to. The application that can see it depends on how it sets up and calls the application. Now this implementation is weak though and can get away without decoding both bytes from the first server response. We should look at how many servers have been chosen and *set* how many servers we can use and those are the ones that send back the appropriate bit.
Now that this is in the public we should improve the implementation for SPDY.
Sorry, I didn't mean to write "data for the first bit (...)" above. It should have been "the length of string"
But basically you're suggesting that the new standard enforce HTTP/2.0-level security and if implemented correctly you mean that if a server wants to use half the length of a web page on the first hop it must encode that bit into a byte, sending a message that is longer than the data it's trying to send? Doesn't sound good.
Well, it is the "lifespan" of the TLS channel after all, and half the page is probably what'll be delivered with most cases. I could easily see some solutions to allow it with more effort, and even then I have seen it disappear before, like not being able to notice that you are being sent only half the page. But why bother with what this standard suggests?
Quote:
I think this won't work outside of browser because the header actually defines server. I would bet many clients will already have the server in name and will still follow the client RFC. Also the header will provide clues to where the data is.
Well since there's no server-side encryption, I would bet most people would use an HTTP/2 server name, but clients would not have that information.
True, but since I don't see how this could work as a client, I have no idea what this concept is that you're describing.
I would think you have to deal with the headers the first time you call the server. As for the protocol: send as long as possible up to the first handshake
Comments
Post a Comment