I think this is dangerous. In the worst case, confidential information could be contained in the mere existence of the binaries. For example, suppose Intel has invented a new and secret floating point unit, the "xo4" array processor. Intel gives you a program called "xo4.exe" which they tell you will dramatically improve floating point performance when you run it over your source code, and after doing so you will have to link to the "xo4.dll" file which they also supply.
I can easily foresee a number of secret capabilities which the CPU might possess which would be revealed by the existence of binaries, especially as these binaries will likely have to be documented to be useful. For example, suppose that you get a binary called "powerpc.exe" and they bluntly tell you that this enables hardware emulation of the PowerPC CPU, but that the existence of this capability is a strict secret. As I said earlier, given the "Appendix H" strangeness where Intel tried to make a trade secret out of the "VME" bit, anything is possible. It would be really hard to make something like WINE work, to cite an obvious example, if you are prohibited from using the "VME" bit because you are not allowed to disclose its existence. -- Mike On 2000-05-26 at 02:20 -0700, Joey Hess wrote: > Hm, as I read this, said confidential information is all contained in some > binaries. > > If you don't look in the binaries, and you don't distribute the binaries, or > distribute code linked to the binaries, have you not complied with all their > obligations without infecting yourself?
