Generally everything aounds awesome. One detail that made me facepalm though,
> DATAGRAM frames do not contain a multiplexing identifier, unlike QUIC's stream IDs.
This sounds like the stupidest plan ever. There's a nice multiplexing protocol, but as soon as you want unreliable or datagram messages, you need to re-invent a new application level protocol for routing datagrams to their specific consumers? This is the dumbest shit.
There should certainly be different channels of datagrams flowing in quic. How we would think for a second anything else makes sense is is uncomprehensible. What the frak?
And indeed, MASQUE right away hits this inconcievably stupid transport limit & has to invent their own "quarter stream id" to figure out where given datagrams are coming from or going. Anytime a unreliable protocol needs to stream over quic, it too is going to have to invent it's own similar transport protocol wrapper, rather than be able to simply just work directly atop quic. This is such a vastly stupid thing, criminal negiglence in protocol design, is a vast world of infinite suck. Quic datagrams, you screwed up! Wicked bad!
QUIC datagrams not having a stream ID was a compromise, which is why the H3-DGRAM draft exists to add them. Any other protocol can use cite and use H3-DGRAM even if it itself is not using HTTP/3.
[+] [-] rektide|4 years ago|reply
> DATAGRAM frames do not contain a multiplexing identifier, unlike QUIC's stream IDs.
This sounds like the stupidest plan ever. There's a nice multiplexing protocol, but as soon as you want unreliable or datagram messages, you need to re-invent a new application level protocol for routing datagrams to their specific consumers? This is the dumbest shit.
There should certainly be different channels of datagrams flowing in quic. How we would think for a second anything else makes sense is is uncomprehensible. What the frak?
And indeed, MASQUE right away hits this inconcievably stupid transport limit & has to invent their own "quarter stream id" to figure out where given datagrams are coming from or going. Anytime a unreliable protocol needs to stream over quic, it too is going to have to invent it's own similar transport protocol wrapper, rather than be able to simply just work directly atop quic. This is such a vastly stupid thing, criminal negiglence in protocol design, is a vast world of infinite suck. Quic datagrams, you screwed up! Wicked bad!
[+] [-] achernya|4 years ago|reply