gistfile1.txt

[16:02:37] <exarkun> If it could be perfect, you could argue just make it a perfect proxy and forget about it
[16:02:52] <exarkun> Since it can't be, should it be a really good proxy that just fails at unpredictable times?
[16:03:07] <exarkun> Or should it be a very limited proxy that's obviously incomplete and doesn't trick anyone into thinking it's better than it is?
[16:03:41] <exarkun> which comes back to - what's accomplished by proxying more than the endpoint implementations themselves need?
[16:03:59] <exarkun> (not a rhetorical question, if we know what the end-user value is maybe the trade-offs become easier to weigh)
[16:06:51] <exarkun> Also, endpoint.connect() fires with the actual protocol, not the wrapper - right?
[16:07:12] <exarkun> So there's actually no reason to proxy arbitrary application interfaces
[16:07:27] <exarkun> Only the interfaces between the transport and the protocol
[16:07:50] <exarkun> That's different from ProtocolWrapper - which ends up exposed to application code.