Of course it's possible to build unsubvertible search algorithms. I present Algorithm A, which returns all known web pages in alphabetical order. :)
The question of whether it's possible to build unsubvertible search algorithms with performance equivalent to the current ones (for whatever definitions of "performance" you find useful and interesting) is a more subtle question.
My feeling is that, even taking aside such ridiculous examples as the one I proffer above, removing features that can be subverted will lead to weaker, less interesting models. (I grant you that there are some interesting problems in adapting/constraining such features so as to make their subversion either difficult or expensive without adversely affecting good actors.)
no subject
The question of whether it's possible to build unsubvertible search algorithms with performance equivalent to the current ones (for whatever definitions of "performance" you find useful and interesting) is a more subtle question.
My feeling is that, even taking aside such ridiculous examples as the one I proffer above, removing features that can be subverted will lead to weaker, less interesting models. (I grant you that there are some interesting problems in adapting/constraining such features so as to make their subversion either difficult or expensive without adversely affecting good actors.)