IANAL, but my reading of the quoted article 10.1, particularly the phrases "...that the form in which a program is, whether in source or object code, does not affect the protection..." and "...obligation to protect computer programs as literary works means e.g. that only those limitations that are applicable to literary works may be applied to computer programs..."
The first phrase appears to apply to efforts such as re-factoring in that the new form of the code does not spawn a new copyright, but instead the old copyright applies to the new form.
The second phrase appears to preclude other forms of protection by declaring that only limitations applied to literary works may be applied to computer programs.
Together with the 50-year term (precluding shorter terms for "applied art" in the last sentence), this article appears to extend the length of time a piece of code can be protected (from a renewed 20 year coverage of 40 years to 50 years without re-application), and also implicitly broadens what counts as a single work from one file of code to all forms that do the same thing. So in other words, any new invention in software that gets applied is not only protected for longer (and under copyright infringement), but also under a fuzzier definition to only include other "forms" of the same computation. I.E. any code that returned the Fibonacci number for any given number n, regardless of the implementation, could now be called the Fibonacci Code, and would be protected as a work by the first implementer (again, any language). This would appear to be in line with the idea that under traditional copyright, the work of the adaption of a creative work also belongs to the copyright holder. So, not only does this make the Fibonacci Code in Fortran owned by the same person as the Fibonacci Code in PHP (regardless if the author may only know Assembly), but also any other derivative work of the Fibonacci Code . Such as one that outputs the numbers as it runs, one that caches an array of values to save on processing time, one that calculates the number by looking up in a table or by recursive calculation, all of it is generally reserved for the copyright holder.
Fibonacci was used as a work old enough to be beyond even this treaty's coverage, but if this could apply to something more recent, say the code that enables spawning new threads of arbitrary code. This interpretation would mean 50 years of copyright claims to any implementation of thread-spawning code (machine up through scripting) by 1 individual (or, hey, the "other" kind of individual - the corporation), and damages/fees as high as the absurdity seen in recent RIAA cases on 1 song, let alone many code translations in use on an idea like thread spawning for arbitrary code.
But, again, we're looking at how far you sub-divide the subject matter in question. A copy of Apache? Definitely whole-sale copying. Using some of the same libraries as Apache in a manor that's similar to one of the functions at Apache's core? Not so much. If one divides "a computer program" to only mean anything one can order the operating system to run and it'll perform the stated task, then the broad definition doesn't affect languages that don't compile to Assembly such as Java. If one divides it too far to mean any function or subroutine which a computer can run, then you basically have software patents with broadened scope (across all languages, and right to create derivatives reserved) and lengthened protection.
This nightmare scenario is essentially over-extrapolation from the starting point above to the legally absurd extreme. I recognize it as such, but not every judge making calls on a related case will see this side when it's spun the right way by the prosecution attorney. The end of software patents may be desirable, but I see the sentence about the code's form not affecting the protection as a very slippery-slope in the wrong direction. Be careful what you wish for. The devil you know may actually be less tyrannical than the devil you don't.