I noticed that ibrowse was polluting the virtualmachine with atoms when passed bogus URLs.
The functions ibrowse_lib:parse_url calls list_to_atom to get the URL protocol but if it's not a well formed URL it will pollute the node with useless atoms.
I see 2 possible solutions:
I suppose I could use list_to_existing_atom, but that is probably only supported in newer versions of Erlang/OTP. Thanks for the bug report. I'll try and come up with a solution.
Ok, I've decided that list_to_existing_atom is a safe thing to do. This will be part of the next release of ibrowse.