Thou shalt have no other gods before the ANSI C standard 1638

Well, then the C standard does not agree with you about what "two's-complement" means. Since this is about a C program, and the documentation is there to help readers understand unambiguously what is happening and what property are buttumed, then a clarification may be in order. At least for the poor soul for which "two's-complement" does not automatically imply sign extension upon right shifting.

What "packet" ? When I look at your paper, it speaks about messages whose length "can be any nonnegative integer". It seems to me that this MAC algorithm should work on just any data message that the computer can manage. Files bigger than 4 GB are not uncommon nowadays. Since the sample implementation requires the whole data to be in memory, that implementation should at least be able to process any message which does fit in memory. Hence the "sizet".

The existence of an Alpha machine running Tru64 is hardly speculative: at some point I had one in my office. And a 128-bit integer type makes sense: such a type is necessary in order to take advantage, in pure C, of the umulh opcode, which computes (a*b)-(2^64) where a and b are 64-bit integers. Other vendors may follow the trend.

--Thomas lovein