I don't believe in ideological freedom - freedom defined on an absolutist basis. Freedom that doesn't maximize its propagation is the kind of freedom that doesn't perpetuate.
Therefore, an ideological freedom that allows anyone to reduce the freedom they received when passing on to others reduces freedom for the wider body of recipients, even though it maximizes the freedom of the first recipient in the chain.
A reduced set of freedoms that is maintained for most recipients maximizes the freedom of the total of recipients, while an absolutist freedom can be cut off absolutely at the first transfer.
Software is not open by default - this is copyright law. Without an explicit licence, no rights are granted, other than provided under the Fair Use Doctrine. So the default state is that no rights are propagated.
So GPL does not "steal" code, and is not "viral", because without a licence, people can't use the code in the first place. You deliberately have to create a derivative work from the original for the licence to apply to your derivative work.
If your work is not a derivative work, the licence does not apply, no matter how you use it. You'd have to be incredibly dense, incredibly incompetent, or both, to accidentally create a derivative work that causes your work to be "stolen" or "infected". Abstinence is 100% effective in this case. If you try to argue "but that means it isn't really free", remember that the code didn't have to be open to begin with. It is much more free than not-free.
BSD/MIT licences don't explicitly require maximal propagation of freedoms. So freedom can be cut off at any point along the chain.
GPL does not inhibit commercial use. In fact, when viewed through the right lens, it has many properties that make it attractive for commercial use.
For example, the fear is that by opening up code in general, there can be no commercial gain, and competitors can piggy back off the originator's investment. However, it must be realized that the competitor cannot use GPL code without opening up their code. This is in contrast to BSD where competitors can piggyback off others' investment and get a free leg-up.
Even more worryingly, anyone can take non-GPL code and make incompatible changes and potentially kill off the original software if it becomes more popular. Embrace, extend, extinguish is very much reality for unconditional apparent (but not essential) freedom.
Of course, all the traditional reasons for preferring open development also apply. Further testing, fixes, improvements, and features are practically free - moreso for GPL-style openness. A contributor may invest a lot of effort into one aspect of the code, but will not have to invest any effort into code that other people contribute. Once it gets to a critical mass, there is a feedback loop, as more contributors means the attractiveness of the code goes up (eg, more features, fewer bugs, greater range of capabilities), meaning more contributors, and reduced overall investment required for everyone involved.
There is no worry that a competitor can introduce incompatibilities because they must licence their derivative work under GPL.
Copyleft is a pragmatic approach to open source. It recognizes that the real world exists, and in the real world we have copyright law. Copyleft works by leveraging copyright law.
It also recognizes that software does not come without effort, so paying for code with code is a very fair transaction. A derivative work would not exist without the original work, and creating a separate original work would be a waste of effort. And the fun part is, if you do not create a derivative work, then you don't even have to pay with code.
An ideological approach would be to ignore copyright law, or call for its abolishment, and pretend that we live in a world with no dark corners and malevolent tendencies. An ideological approach would view the barring of the propagation of freedoms as a freedom, whereas most legal systems around the world, and common sense, realize there can be no such thing as absolute freedom.
Software patents are toxic for the software industry, end of story.
Software patents are unreadable and too wishy washy. In software, all that matters is the code. Software patents stifle innovation by being way too broad. If a piece of software can accidentally infringe on a software patent, it should mean the software patent was not novel or non-obvious enough to begin with.
Tivo-ization goes against the spirit and the propagation mechanism of copyleft. Code is pointless if it cannot be used to create something that works. Even if the code were to be written in some language with built-in formal proofs, it eventually has to run on a real machine that may break hidden assumptions about the hardware in the code.
On a more pragmatic concern, with the rise of Spectre and Meltdown, being able to verify code in real world conditions has security implications.