There's a difference between patents that happen to cover software and a carved-out subsection of patents designated specifically for software by law. We currently have the former - Niley believes we should have the latter.
1. There is nothing inherent about "software" that suggests you can't create a patentable invention with it, therefore statements like "software patents shouldn’t be allowed" are wrong, or at least equivalent to "patents shouldn't be allowed".
2. Software has certain properties that result in undesirable side-effects or ambiguities under the patent system at present, therefore we should adjust the treatment of patents on software to mitigate those effects and clarify those ambiguities. An example would be reducing term lengths to something more compatible with the pace of innovation.
I don't understand, how does that work? Let's get rid of the word "software" for a moment. Say I think that patents involving ducks shouldn't be allowed, for whatever reason. There's nothing inherent about ducks that means you can't make a patentable invention with a duck. But that by no means implies that my stance against duck patents means I'm against patents as a whole.
I would also point out the other possibility in what I said before: that statements like "duck patents shouldn't be allowed" are wrong. Maybe "wrong" sounds too absolute, so substitute "incomplete" if you like. "Some duck patents shouldn't be allowed" is a different proposition, and probably closer to what most people actually think, but Nilay is specifically addressing the absolutist rhetorical frenzy on you see on sites like Quacker News.
As for "some", you may be right, but I'm hard-pressed to think of any worthwhile software patents. Really, the fix for software patents is probably to reduce the term and vastly increase the requirements for novelty rather than eliminating them altogether, but on the other hand I really don't see much downside to simply disallowing it completely. Maybe I'm missing it, but patents don't really seem to be a requirement for software innovation.
Put another way: What's so special about "not-software" or "not-duck" that excludes it from the issue you have with software or duck patents?
That's pretty much my point. The original construction assumes that not only is there nothing special, but that the person arguing against software (duck) patents thinks that there is nothing special.
My real complaint is phrasing it by saying that a person who argues against software patents must be opposed to all patents, when it would be much more correct and much less insulting to say that this person should be opposed to all patents. One reduces the other person's argument to an absurd degree without allowing for any possibility that they might have a point, and the other just states why you (or whoever) think they're wrong.