3 ms·
It probably is. So what is wrong with say MurmurHash that you have to hack your own?
by AstralStorm 9y ago
It probably is. So what is wrong with say MurmurHash that you have to hack your own?
- jimktrains2 9y agoYeah, it's one thing to create it for fun, but why publish it as a library for general consumption?
- rjeli 9y agoI strongly disagree with this idea - code should always be released! If it's good, we can avoid rewriting code and have more options for hashing, and if some people point out problems with the code, the author can learn from their mistakes and fix them if possible - they wouldn't want to use it in their private code with the mistakes. I think this thread is more negative than the author deserves. The method of hashing is novel and could possibly lead to exploration in the area, and if not that, where else could the author post to get feedback from the author of murmurhash?? I think the only problem here is that they advertise it as production ready/universal without evidence or rigorous testing, rather than "hey, here's a cool hash I came up with."
- AstralStorm 9y agoIndeed. That claim requires hard statistics on performance on multiple platforms and statistics of the hash function itself. The floating point hash might be useful on a CPU for example.
- jimktrains2 9y ago> I strongly disagree with this idea - code should always be released! There is a difference between releasing code, say on github, and putting it out there like it's ready for use in a production system, like in a package manager. I only said the later shouldn't be done, not the former. Rereading, I guess that wasn't exactly what I said. I can't seem to edit now, though.
- 19eightyfour 9y agoThanks for sticking up for me, rjeli, I really appreciate that!
- 19eightyfour 9y agoI like making things. I want to have an impact. I want to do something new. I want people to try and use my stuff and tell me about it. So I'm achieving that today. Make your own hash function. It's not that hard. Most non crypto hash functions are just something similar to ax + b mod p. And then there's plenty of room left for exotic ones. There's room for yours, too. Personally I've for a long time really liked making RNGs. Many people like making them actually. For me, I always liked the feeling of getting something simple that i made create so much uncompressible apparent entropy out of so little. And i enjoy the challenge of passing the test suites for randomness. And i enjoy the feeling of making tiny mixing functions. I like the diagrams of mixing functions and ciphers. I like the aesthetics of writing that code, both that which is similar to the main habits of many ciphers, the xors, the shifts, as well as coming up with something new. A lot of people like making hashes, and RNGs, and there are a lot of them, and a lot of good, useful and interesting ones. All made by people no smarter than me. I'm fine to get up in front of you all here and try and do something i like, and fall and succeed, and learn. I just enjoy what I'm doing. I'm creative and can come up with new things. I like doing it. Coding is one way i can express that creativity in a way that satisfies me. And i feel more satisfied when i share what i make it with others, and i have more fun when i take it more seriously. I like a challenge. And i think i can do something good with this space. And i want to. And I'm fine to fail too. I like to succeed. And I'm fine to fail. I just like doing it. I like doing more than i like sitting on the sidelines and commenting on other people doing, and secretly wishing i was on the field doing something too, instead of in the stands just having opinions. That always felt fake to me... Because if i secretly wanted to do what someone else was doing, I wouldn't get jealous, I'd just say to myself, how do I get that, too? Because, why not? So other peoples achievements inspired me to do, too, if i liked the same thing. If didn't like it, I wouldn't do it, but i could still enjoy it vicariously. So that's why I do stuff and share it with others and take it seriously. Hope that answers your question.
- 19eightyfour 9y agoI answered the why below. Also, i didn't feel there was anything really "wrong" with others. One motivation i had specifically apart from that paper i linked asking for a different way, was i was bored of seeing the same hash and type of code in hashes and ciphers again and again. And i had made a successful RNG, using a very simple construction, in a similar genre to this same regular genre, using the traditional operations like shifts and xors, and i felt i pretty much had nailed that type of construction to my satisfaction. So i wanted to do something different. A hash with a different aesthetic. A hash that, when you look at the code, it looks different to others. I didn't like that i just had success in this one genre of hash or rng code, and i wanted to do something just using arithmetic operations. It annoyed me that i had never really been able to find, or make, a good hash or rng without using shifts or xors or the same boring multiply stuff. When you see enough of them, it's all just derivative... Changing constants and so on...And the choices, this xor instead of that plus, are basically arbitrary because the space of possible algorithms using similar constructions is so huge, they're really not so different to each other, just more points in the same hyperplane, and there exist still so many other algorithms to find that way, a computer could basically fill in the rest of the dots. So instead of trying to discover a new point in parameter space of constructions like that, i wanted to get back to something more pure, that was more about the essence of a different construction, rather than a few more tweaks to the same type. Also, not just pure, but short. So many lines of shifts and xors are unnecessary...i know from deep and long experience, you don't need that much code to make so much good apparent entropy. I proved that by making a good rng, with a very simple construction. And I've written code like the sea of hashes and mixing functions hundreds of times... And really... It is like, change this shift, change this constant, change the construction, tweak until it works... But it's ask basically the same structure... It's just so boring after a while. I'm trying to convey to you the feeling that motivated me to do it, since that is what you asked. Hope it comes across clearly. Finally, i don't want this to come across as only a condemnation. I draw a lot of inspiration, ideas and information reading other people's ciphers, hashes, and papers. It's just, I don't want to copy the same style always. I want to diversify. I have a good mixing function in the traditional genre ( dosy, also on npm ), and I wanted to see if i could construct a good mixing function in a new aesthetic. And i have. Hooray! Tl;dr - aesthetics. I felt bored with what i had seen and done. I wanted to discover, make and see something new.