Closed ccutrer closed 7 years ago
The underlying proposal herein is certainly a useful change.
I have a few concerns as proposed, which prevent me from immediately merging:
I've pushed up an amended commit that addresses 3 and 4.
Thank you :)
re.
I should think if @backlog.size > @batch_size || @backlog_bytes > @batch_bytes_size
should be a fairly easy to maintain condition pair. I think I'd like for us to go this route, as it saves any user surprises. The other alternative is to introduce a separate batch class, but I don't think we're at sufficient complexity to warrant that yet. We can pick an arbitrarily large value for the default @batch_bytes_size
.
👏 🉐 I parry that if we're not considering ideal worlds then we should do as DNS does, and pick 512 bytes, making space for swolen ip headers. ⚔️
I don't think we'll reach a perfect answer for 2, but let me know what you think.
Thank you for your thoughtful responses, patience and contribution!
Released, too!
otherwise it's impossible to properly fill a UDP packet, without risking going over, due to varying lengths of messages