I agree with RMS on many points, but I think you're misguided:
> I've always used the GPL for licensing my open source code
ITYM "free software"
> specifically because it ensures that it stays open and that improvements are given back.
No, it does not. Any software not being redistributed does not have to publish its improvements. And even if it does publish the improvements, they don't have to be made accessible in any useful way (for example, just a source dump, rather than a version control repo). And even if it is made accessible in some useful way, often the projects don't end up contributing back upstream (Hello Ubuntu!)
> On one occasion I've been persuaded to use another license so that $CORP could use the code (http://gmsl.sf.net/) internally and I see that the non-fanatical position helped $CORP but I'm not sure it helped the greater cause of open source
They still can. You're confused about licensing and what GPL does.
HOWEVER, I do resonate with your concerns. I think the world needs to start making programs which are NOT "free" according to RMS, in that if someone uses the program, and he improves it, he has to give back to the general population. How to execute this properly is another question, but I really want to be able to release code which is a public good, rather than something everyone can take, without ever giving back.
The Affero license sounds closer to what you want. It doesn't stop someone from using the code internally without releasing improvements, but it does stop a company from building a network-based product on top of your code without releasing their improvements.
Beyond that, though, it sounds like what you want is actually a bunch of mandatory process bureaucracy. You seem to want everyone to be required to contribute to the original repository, but that's a maintenance nightmare. This would pretty much prohibit forks (pushing changes upstream from a heavily modified fork is non-trivial), and add senseless work to lots of derivatives (e.g. some project B uses a tiny piece from A, but now has to attempt to send all changes upstream, no matter how irrelevant).
Ubuntu doesn't contribute to upstream? What do you mean?
The limitations you point out on publishing changes are nits imho. It's far better than nothing, and social pressure has been successfully used in many cases to make sure proper SCM logs are sent.
> I've always used the GPL for licensing my open source code
ITYM "free software"
> specifically because it ensures that it stays open and that improvements are given back.
No, it does not. Any software not being redistributed does not have to publish its improvements. And even if it does publish the improvements, they don't have to be made accessible in any useful way (for example, just a source dump, rather than a version control repo). And even if it is made accessible in some useful way, often the projects don't end up contributing back upstream (Hello Ubuntu!)
> On one occasion I've been persuaded to use another license so that $CORP could use the code (http://gmsl.sf.net/) internally and I see that the non-fanatical position helped $CORP but I'm not sure it helped the greater cause of open source
They still can. You're confused about licensing and what GPL does.
HOWEVER, I do resonate with your concerns. I think the world needs to start making programs which are NOT "free" according to RMS, in that if someone uses the program, and he improves it, he has to give back to the general population. How to execute this properly is another question, but I really want to be able to release code which is a public good, rather than something everyone can take, without ever giving back.