[go-nuts] Best practices for long running servers

Hi all,

I've been working for a while on a Go implementation of Twitter's Snowflake
unique ID generation service.

You can check it out here: https://github.com/erans/gosnowflake

I would love to get feedback in general and on the code and styling as this
is my first "real" Go based project.

I was wondering what is the correct way to implement a server in terms of
error handling. Do the "Serve" command needs to handle everything and panic
if things go really wrong (memory issues, etc).

I've recently added Thrift support to it and when I forgot to correctly
implement one method the server blew up. Is there a way to catch it in a
more graceful way?

Thanks,
Eran