Squid blocking.IO requests after upgrade about engine.io

[rauchg]

Very interesting. I wonder if trying the upgrade to a different host (eg: upgrade.yourdomain.com) would help avoid this situation

[martintajur]

Allright, I checked my own code more throughly once more and discovered I had the handleUpgrade method bound to httpServer twice which was wrong and caused the upgrade not to actually happen. It's highly likely that this also caused Squid to go nuts.
I fixed my code and I will go and visit that Squid-enabled network soon to see if it works now.

However, I also discovered that when I was actually able to go through the upgrade (using Chrome 18), it produces a 10+ seconds lag each time right after the upgrade before the requests actually start moving. Almost like they get queued somewhere internally and then all of a sudden they start to move, or almost like it takes 10+ seconds for any websocket traffic to actually start flowing... Is this known behaviour?

[martintajur]

Allright, I checked my own code more throughly once more and discovered I had the handleUpgrade method bound to httpServer twice which was wrong and caused the upgrade not to actually happen.

Btw, the reason I had this bound twice was that I didn't pay enough attention to the fact that when attaching Engine.IO to httpServer, it creates the handleUpgrade binding automatically. Perhaps it would make sense to build a simple check to avoid having double bindings?