The argument here seems to be that when GnuPG's implementation, or the original standard, has flaws, those flaws should be seen as inherent limitations of the use-case, rather than flaws in the implementation and standard. And with GnuPG, that argument seems to be used to justify having it behave the same way it always has, which leads to dangerous situations.
That PGP handles armor, escaping, and comments badly, and clients handle display of the signed text badly, do not seem like they mean that the concept of cleartext signatures are inherently flawed.
