Sven Schüle wrote:
Using Don's similarity test is not valid to prove that two programs are "identical".
Not identical, nothing in physics is identical, self-similarity in this test is not 1. Don's similarity utility is _very_ useful in detecting closely related or extremely closely related engines to the point of an almost pure clone. One has only to check for noise (error margins). His first version was flawed, the last one is perfectly adequate.
It was not intended for this purpose, I may assume that you know that.
No, I don't know that, and I am using it not thinking of how it was intended, I know how it works, and use it to my pleasure.
Surely you can show that two programs _behave_ very similar. And in case of Strelka vs. R1 this is no big surprise, since Strelka was explicitly designed for exact that purpose.
My point remains that Strelka can't be used to conclude anything about Fruit-R1 connections.
Wrong. You gave examples A --> B. I gave A = B. Now let's take a real world approach.
By all accounts, Vasik's, Uri's, etc. the main part of Strelka 1.8 behaves like Rybka 1.0 Beta. By many accounts Strelka 1.8 is pretty (not very) similar to Fruit 2.1.
Strelka 1.8 is 90% Rybka 1.0 Beta and 10% something else.
Strelka 1.8 is 70% Fruit 2.1 and 30% something else.
Then, in the best case for your denial, Rybka 1.0 Beta is 66.6666% Fruit 2.1. In the worst case for you, Rybka 2.1 is 77.7777% Fruit 2.1, more than Strelka 1.8. It means that if Strelka 1.8 is related to Fruit 2.1, then Rybka 1.0 could be related a little less or a little more than Strelka to Fruit 2.1.
I hope the relation between Rybka 1.0 Beta and Fruit 2.1 is now clearer for you.