Do we need the [algorithm] tag?

Question

I just found the algorithm tag, our most-used tag without a tag wiki (excerpt).

It has 16 questions (7 asked in the last 30 days), and I can't seem to find what this tag indicates (for writing it into a tag wiki).

Most questions on Crypto.SE are about some algorithm, this is not really specific here. The questions here are of these types:

1. What algorithm can I use to do X? (convert between elliptic curve operations, generate prime numbers (another one), find discrete logarithms, factorize large integers, calculate primitive roots, decrypt only after some time, sign the current time)

2. I need some detail about algorithm X (RSA)

3. What algorithms are used in X? (PGP)

4. Attacks on algorithms (some code which implements a cipher, timing attacks)

5. For what algorithms I need/can use X? ("real" random numbers)

Looking at the numbers, only retaining type 1 would be worth to consider (then we could look if there are more questions of this type not yet tagged). But even this is dubious: Would there be people following this tag (for now, 0 followers), or searching questions about "is there some algorithm which does something", no matter what?

So, my proposal is to remove this tag completely (without replacement).

Any opinions?

Answer 1

So, my proposal is to remove this tag completely (without replacement).

Any opinions?

I agree. Algorithm is a bit like having cryptography on a site about... cryptography. Essentially, any implementation probably uses an algorithm of some sort. So going through your example questions:

1. Such questions should use appropriate tags for the type of problem domain they're interested in, like time-based-crypto or large-primes or whatever it might be. We can always fix this if the OP is unsure. looking-for-algorithm would be a meta tag about the question in this scenario.
2. Details about an algorithm are probably covered just with a tag about that cipher or implementation, e.g. rsa or pgp and perhaps the specific detail the OP is after, e.g. differential-analysis or preimage.
3. Again, as for point 2. Perhaps implementation?
4. Attacks are cryptanalysis?
5. This one strikes me as 2) i.e. "how do truly random numbers actually work in an implementation then?" It's a little bit more abstract and perhaps there's a tag for that. I feel like there should be, but algorithm it isn't.

In short, I agree. Feel free to edit/fix my examples if they're a bit off, I'm currently on the move and just checking in.

Answer 2

There is room for a tag which says the opposite of protocol-design and that could be algorithm, except that this tag already exists under the name primitives -- that's the correct name, but I guess it may be a bit too obscure (it has been used only once). Maybe we want to rename algorithm to algorithm-design (and remove it from some of the questions which currently use it) ?