Luke Gorrie

An extreme example of what Andy mentions is lukego/blog#8.

@mraleph hinted that there _may_ be a simple way to avoid the overhead of side-traces rejoining the parent trace. I posted a bounty on LuaJIT via BountySource in case anybody...

@capr Short answer: that ship has already sailed :). Long answer: I thought seriously about this approach a year ago, see [snabb-devel post](https://groups.google.com/forum/#!searchin/snabb-devel/torch7/snabb-devel/2kNiyRvIl5g/oxhbGpfU5lcJ). I understand that it makes sense for...

@capr well that is my off-the-cuff response anyway. Since this PR is all about looking at things with fresh eyes we can see if other people see things differently too.

> what a can of worms I opened :) I understand the need for the metatype early-binding hack now, although TBH I would still cache on an as-needed basis, close-to-the-loop...

Alternative idea: we could put all of our modules directly into the global namespace instead of a `snabb` module. Here is how the example above might look in that case:...

(Note: The core modules are already in the global namespace today, e.g. `packet`, but people don't use them that way for fear of expensive lookups when referencing the module on...

@szbnwer Here is a random license compatibility diagram off the internet: The most important transition in my mind is that Apache cannot be merged into MIT. This could become a...

Here's my braindump to kick things off. Fight me :-) The long-term target group for RaptorJIT is people who prefer dynamic languages and want to do high-performance system programming. Suppose...

Yes, `proto` was added to upstream LuaJIT very recently, see reference to the relevant commit at the end of the PR description. Could likely reformulate the code to not depend...