I think the only question you raise that’s significant here is about multiple versions. If your objective is to support multiple versions of the same assembly then the clear winner is to use the GAC. That’s what the GAC excels at.

Your point about disk space and performance are correct. If your only objective is a slight performance increase or a slight savings in disk space then there’s really no compelling reason to use the GAC.

Carey

How much a performance hit is it to have your dependencies looked up from the PATH? is it significant?

Now that disk space no longer seems to be at a premium, what is the real gain (other than the slight improvement in performance maybe) by keeping shared assemblies in the GAC? If I need to maintain two versions of the same assembly in the deployment environment, it seems much easier for each application to have their own versions than a global cache.

What benefit that the GAC provides would compel me to switch to the GAC ?

I’ve not read every single line of the current bill. I do support a single-payer option which seems to be the big sticking point for many. $900,000,000,000 is the proposed cost for a 10 year span. President Obama has stated that we have to allocate the money up front. What I don’t trust is that according to the administration the insurance and medical industry has promised to offset that cost by $600B, leaving $300B to the tax payer over ten years. I don’t trust the insurance and medical industry to do their part though. I think (and have no proof) that those industries feared that the proposed plan might make and had to be on “the winning side” in order to strengthen thier political position in the future.

But let’s assume that the $600B is real, that leaves $30B/year to cover. That’s a pretty small amount for a tremendous benefit given what we currently spend on.

As it stands now, I think the discussion is moot. I think single-payer is dead.

Damian Conway gave a talk at OSCON called “Temporally Quaquaversal Virtual Nanomachine Programming In Multiple Topologically Connected Quantum-Relativistic Parallel Timespaces… Made Easy!” (just over an hour long, but worth it as always) in which he related a story, around the 37min mark, about John Ellis losing a bar bet that forced him to work the word ‘penguin’ into his next paper. If I get another shot, I’ll go for ‘monad’!