Author is a Lungfish fan, which I appreciate, but using a lot of examples of bad icon design does not quite prove his point. He mentions himself that icons have the power to hold way more language than can fit verbally in the same space. Once a user learns an icon — even a bad one — it can save mental overhead on repeat uses.
Good icon design is not easy, and all these bad examples by big software makers definitely drive that point home, but what that says to me is "don't overestimate your skill at making understandable icons."
So if you can at least pull that from the article, that's a good message. But I wouldn't want everyone to suddenly feel inadequate and give up on iconography altogether.
Yeah, I agree with your sentiments here, particularly: "don't overestimate your skill at making understandable icons."
Usability is important, to be sure, but graphic appeal also has its place.
Well known companies are training their users to recognize icons, which become part of their UX and in turn their brand. These standards often see wide adoption and become widely recognized.
This provides interface designers another tool to make cleaner, more attractive pages.
A confusing icon can certainly do more harm than good, and text rollover hints or just plain text labels can often be a better choice.
The author is not saying that graphic appeal doesn't have its place. However usability will always trump graphic appeal as the end goal of an interface is always directly tied to an action.
When Apple, or another large company introduces a new icon, they are in essence training their customers on what it means. This could have a negative impact on UX out of the gates, but in the long run can make UX better.
Every icon we have ever used, we have had to learn its meaning.
Brand and marketing concerns can and do trump usability in certain circumstances.