Re: [quicwg/base-drafts] Connection migration optional? (#1271)

[erickinnear <[e-mail]>]

Yes, I think we're in agreement that it will be useful to have a transport parameter, I just want to make sure that we're careful not to back clients into a corner where they cannot behave correctly due to influences outside of their control. A client that is behind a NAT might be surprised that the server thinks it's behaving incorrectly if it wasn't aware that some portion of the routing/NAT binding changed significantly. 

That's mostly a wording issue, I think, not a problem with the concept itself. 

It almost seems like it could be as simple as using the same terminology that we used previously ("local address" vs. "from an IP address"): 
```
The server does not support connection migration (link to the other section). 
Clients MUST NOT send any packets, including probing packets (probably 
use the real name from later), from a local address other than that 
used to perform the handshake.
```
It might be interesting to add, either in the definition or later where the second reference is made, a note that clients may wish to be more aggressive about using keepalives if connection migration is disabled. 

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/1271#issuecomment-397517679