confirm licensing

[Licenser]

Hey @zwegner, I ported your code to rust and included a LICENSE-UPSTREAM with your original license. I'm always a bit unsure of how to handle licensing with ports so I figure it's simplest to ask if you're fine with that :).

[zwegner]

I guess my only concern would be whether I would still have partial copyright on the end result. It's a bit of a tricky subject here, where the code is somewhere between a mechanical translation and a clean-slate reimplementation of the algorithm. IANAL, but I think here, where there's still a good amount of my original text left (mainly comments), I would still have partial copyright. In that case, no need to add a separate upstream license file.

Dual licensing MIT/Apache should be fine. I'm not aware of all the ramifications, but I mostly just wanted attribution in code that use my library.

On a side note, I will note that there's an out-of-bounds memory access bug in my master branch code that is replicated here (it's fixed in my wip branch, but not rebased+cleaned up). In the main loop, the condition is while offset + V_LEN < len, but the unaligned load of shifted_bytes is from the pointer offset + V_LEN - 1, which accesses bytes up to address offset + 2*V_LEN - 2. It's not quite trivial to fix this, since you need to do one of these:

• have two vector's worth of validation in the remainder code after the loop
• load shifted_bytes from offset - 1 before each validation, and have special handling for the first loop iteration
• use an instruction like vpalignr instead of any unaligned loads (this is generally slower in my testing)

[zwegner]

Oh, and also: thanks for doing this! faster-utf8-validator is one of the first projects I'd "advertised" in a long time, and it's really nice to see it getting some traction.

[Licenser]

Thanks for the feedback. To be honest, I was less worried about the legal part as the 'doing the right thing' part :) and it seems we got the same approach there. I'll make sure the app calls out your work appropriately!

It is really nice work! :D I liked the code, it's clean well explained and pleasantly structured.

[zwegner]

It is really nice work! :D I liked the code, it's clean well explained and pleasantly structured.

Thanks! It's definitely gotten worse throughout the wip branch, and my choice of C over C++ is looking less wise in retrospect... I'll let you know when I finish documenting the algorithm changes there.