For almost 7 years I've been working in algorithm creation for new encryption systems. I'll be the first to tell you that my goals are lofty and that it's no small task to make something worth using. With that in mind, is Crypto the place to ask about how implementations work or ask for clarifications on an existing algorithm's process?

Sample questions such as:

What attacks does this algorithm's technique defend against?

What is the benefit of this technique over this technique?

Some questions that may be on the edge of being subjective (but may need to be asked?) might include:

Is there an error in this implementation of this technique?

Will this optimization cause problems?

Basically any question that might include code. What's the limit on questions relating to new algorithm creation or techniques?

If there is a SE site where a such a question is on-topic, this is crypto.SE.

However, past experience (e.g. with Usenet group sci.crypt) shows that there is high potential for questions like "hey I do not know crypto but here is an algorithm [some sort of unreadable code in an obscure language] what do you think of it guys ?", or even "my new algorithm rocks, try to break it [followed by a sequence of a thousand digits]". The sample questions you cite are fine and on-topic, but I would argue for some clear and strict guidelines. We do not want people posting complete specification of their new marvelous cryptosystems, in particular if the specification consists in code in some programming language.

As a basic rule, the question shall be precise and narrow, like "how do I estimate the cost of linear cryptanalysis ?", and certainly not "is it secure ?".

• It's a fine line between acceptable and not. – Corey Ogburn Jul 13 '11 at 19:31

This is primarily a site for experts, so I would say that this is a perfect place for those type of questions.

However, everything needs to be taken on a case-by-case basis.

I agree with Thomas Pornin in general.

Plus, if any such questions are to be asked, I think people should be encouraged to use mathematical notation here instead of a sample code in any specific programming language. Even if the question is based on (to some extent) an implementation in any programming language... Or they should use pseudocode at least.

This site is not for computer engineers or programmers alone, since the core subject here is cryptography in general. I strongly believe the questions should be at least readable by a large portion of the users.