|
|
|
|
|
|
by michaelmior
1090 days ago
|
|
|
> You can reuse the connection but that means queuing requests and hitting the latency as a multiplier. This is partially true. HTTP/2 is capable of issuing multiple requests on a single connection and retrieving multiple responses in a single packet (assuming it fits of course). So while you probably won't get the same benefit as you would with multiple parallel connections, the overhead is likely to be much less than just a simple multiple of the latency. This is especially true if you have a large number of small assets. |
|
|
> HTTP/2 reduces the problem with multiplexing but it still doesn't go away
Even multiplexed you hit initial TCP window limits + all the response header overhead.
And many cases you have dependencies - need the first HTML content to issue the JS request to issue the Auth request to issue the API call to issue the dependent API call...