I have noticed what seems to be a trend of general confusion about what should be done when pure IP networking questions are posted to Stack Overflow. I have collected some very recent examples over the last few days:

To summarize:

Some Users & Moderators dislike theoretical IP networking questions on Stack Overflow and close them:

• http://stackoverflow.com/questions/6479554/the-journey-of-a-packet-from-my-computer-to-the-web-server-and-back (Closed: Off topic)
• How to decode an ARP Packet? (Closed: Off topic, then reopened upon my request)
• When can we see a TCP header without any flag set? (Closed: Not a real question, subsequently deleted by mod team)
• IPSec over WLAN requirements (Closed: Not a programming question)
• About NAT piercing (User deleted the question after it was closed)

Sometimes theoretical IP networking questions remain on Stack Overflow, as suggested by Gilles:

• Number of network connections possible
• How do applications know which OSI protocol to use
• Number of hosts in a subnet
• Relationship between TCP and IP Packets
• How does packet interaction with TCP Selective Acknowledgement work?
• Why might an IP datagram be encapsulated in another IP datagram? (Remains on SO)
• Leaky Bucket Problem (Remains on SO)

Sometimes networking implementation questions remain on Stack Overflow (instead of migrating to ServerFault)

• IPv6 Link Local Address Format
• SNMPv3 configuration (Remains on SO)

Sometimes theoretical IP networking questions get punted to Server Fault:

• What is a VALID subnet
• How is IP Restricted to Forwarding on Spanning Trees
• What is the difference between a gateway and a router

Sometimes DNS and other networking services questions get punted to something other than Server Fault, because of another popular keyword in their text

• One Domain, multiple IPs (Posted to SO, migrated to Webmasters because he was talking about websites, but the the real issue was intelligent DNS responses, not html or httpd)

In this example the OP asked questions about how to use research-grade techniques, but he got punted to a forum that cannot help

• Is there a single ended available Bandwidth Measurement Tool (Migrated SO -> SuperUser)

To Moderators & Question-Close Voters:

My bottom line suggestions for improvement are:

• If the subject is IP networking, please do not merely close the question as off-topic; place it in what you think is the right forum.
• Could someone address IP networking in the StackOverflow FAQ to reduce ambiguity?

I write this because IP networking is my career, and like programmers need SO, networking specialists need some consistent place to get a good answer... and the field is wide open... it is not just generic questions about IP networking or how do I deploy XYZ service... if it was well-done, it could conceivably expand into what is currently dominated by non-StackExchange sites right now... like Cisco NSP, Juniper NSP, and perhaps smaller network vendors which don't have a useful place for users to gather and collaborate.

I'm pretty sold on the value of stack exchange, but it can only be as good as the consistency of moderation for a given subject. With great respect to the mods on StackOverflow, ServerFault, and SuperUser, the consistency of moderation for networking questions seems like it could be improved.

If I am correct, what is a good path to get there? If I've misunderstood something, please set me right. Thanks to all the stack-exhange staff and volunteer mods who have built a fantastic discussion framework for programmers.

BOUNTY GUIDELINES

This bounty will be awarded to the anyone who can demonstrate that the following have been done:

1. Stack Overflow's FAQ has been updated with exact guidelines regarding IP Networking questions. I even have some suggested guidelines:

• Any IP networking question involving host-level theoretical or practical TCP or UDP issues belongs on Stack Overflow
• ALL other IP networking questions belong on Server Fault.

2. The IP networking questions listed above as closed should be either reopened, or migrated to the appropriate site.

Just because Gilles has a good answer to my original post does not mean I would not award the bounty to another person who makes the changes above.