Re:Magic Numbers
by cquintana92 on Thursday April 25, 2024 @12:21AM
(#40153383)
48 bytes which doesn't include the header, which was decided to be 5 bytes because it was "around 10% of the body", making it... 53.
I'd loved to be in the room when those decisions were made. I can almost imagine some arbiter hearing the arguments from both sides, and then, to the horror of everyone, deciding that if they can't agree on a common value, they'll have to meet half-way.
OTOH, skimming the Wikipedia article, it seems to me that overhead in that system was proportional to message length, so perhaps length of 53 bytes was still meaningfully better than 64 bytes.
It's an old story. One side should have renounced their claim rather than live with the abomination, and then the arbiter could have chosen their proposal as the one true packet size. The arbiter's wisdom would have been celebrated for millennia.
That's like a programming language with 0.5-based array indexing.
It's the classic "off by half" index bug.
Re:Magic Numbers
by BizarroLand on Friday April 26, 2024 @02:28PM
(#40172584)
Doing the mental math is a lot harder, too. Any halfway decent computer nerd knows on some level that 2^10=1024, but 1/2^10=0.0009765625.
Doesn't quite lock into the memory banks so easily.
No fence post problems if you just put each post in the middle of each span!
Don't give people ideas https://gist.github.com/MawKKe/1fafc1f4a5eb79ed654cf0567c284...
[1]: https://en.m.wikipedia.org/wiki/Asynchronous_Transfer_Mode