ironsavior/phlegethon

Dynamically manage firewall rules based on peer activity

phlegethon's Issues

Workaround bug in boost::this_thread::sleep_for

On Windows, boost::this_thread::sleep_for() results in the thread sleeping for about double the desired time. This bug should be fixed in Boost version 1.54.

Add option to override default pcap filter

Just in case we don't like it.

Rework network structures so they don't have to be packed

Write routines to construct ip_header_t and udp_header_t objects from input streams.

These objects are currently constructed by copy-constructor by casting a pointer to an array as the object to be copied. Doing it this way requires that the structures be packed (#pragma pack 1), which is less than ideal. This is a very C-ish/hackish method that is perhaps not completely appropriate for a C++ program.

Minimize/compartmentalize dependencies on pcap C library

I'm having trouble making this build on Windows seemingly due to inclusion of <pcap.h> in C++11 source files. It may be a reasonable workaround to reduce all contact with pcap.h down to just 1 source file and perhaps write that file to be compatible with C++03.

This is also an opportunity to rework the callback system so that I don't have to pass a pcap_pkthdr struct, but instead a more C++ish type so we can take some boilerplate time calculation code out of the packet handler.

Need to static link the windows build

standard libs are probably not present for systems without the dev environment.

Infer source host address from interface specification

Allow users to omit the source host address and infer the addresses from the interface specification.

Make it portable to Windows

That would be pretty rad.

Add option to list all network adapters

Interface names on Windows are ridiculous and defy memorization. There should be a command line option to get a list of them or maybe the ability to choose one.

Take implementations out of headers.

Stop being lazy. This needs to be done for better build automation and to keep from polluting source files that include my headers with symbols from C headers.

The Stats namespace could do with some better abstraction

A few key points:

• peer_data_t should not derive from std::map.
• Support an arbitrary number of data points per series
• Support an arbitrary number of series
• Support arbitrary types

Move header loading routines

There's no good reason the Net::load<> functions should be implemented as templates. Think about reworking the namespaces so this makes more sense. Do these functions more rightly belong to the libpcap namespace rather than the Net namespace? Should these functions be implemented as static methods or constructors for the structs, directly?

Need to ponder this a while.