There's a lot of overlap between file extensions and TLDs though. .py, .sh, and .app are some more examples (but a fully exhaustive list would be in the dozens if not hundreds). At some point you have to just treat them as the separate namespaces that they actually are (and not somehow try to block a TLD from being used as a file extension, or vice-versa).

Besides, accidentally resolving a file extension to a TLD is only one of many possible different serious errors that can result from exposing an API that can load files locally or remotely, and thus make network calls that you might not be expecting. Fundamentally you need to fix that API either way.

not to mention .com

Geez, that's by far the best example and it didn't even occur to me. And it's an executable file type. Good call.

I think it's okay. They are different namespaces and people should fix their bugs. It would be like skipping the street number "911 Foo St." because "911" is the emergency phone number and 911 Foo St. is just a regular house. I'm sorry if someone is confused but they shouldn't be confused. They're totally different things.

The biggest mistake we made with DNS was the "shortcut" of implicitly adding the root domain to random strings treated as domain names (turning "example.com." into "example.com"). The file "foo.zip" and the website "foo.zip" wouldn't even be ambiguous if we called the website "foo.zip.". "ndots" also causes operators of DNS servers a lot of pain -- some malfunctioning program tries to resolve "example.invalid" in a tight loop and it balloons to asking for "example.invalid.", "example.invalid.local.", "example.invalid.cluster.local.", "example.invalid.svc.cluster.local.", and then DNS blows up, breaking everything.

> It would be like skipping the street number "911 Foo St." because "911" is the emergency phone number and 911 Foo St. is just a regular house. I'm sorry if someone is confused but they shouldn't be confused. They're totally different things.

Floor 13.

We just never should have allowed filename extensions to have semantic power. Resource forks are far more elegant and you could do simple look ahead checks to verify types etc.

Back in '81 when MS-DOS came out and ascribed "magical power" fo files ending in ".com", DNS didn't exist (except possibly as experiments stuff between a handful of university or military mainframes). It was also an era where the vast majority of users of those .com files were doing so off 5 1/4" floppies (or possibly even 8" floppies), and the very idea of adding "resource forks" to your on-disk data would have been ridiculed as an outrageously profligate waste of extremely valuable disk storage resources.

It's easy to say "we should never have" in retrospect. But you're basically accusing programmers back in the late 70s of having insufficient foresight to see problems that would make their decisions seem bad almost half a decade later.

We also should never have let companies sell cigarettes. Or burn fossil fuels. Or start social networks.

The magical power of extensions actually predates MS-DOS -- both CP/M and TRSDOS did the same thing in the late 70s, and I'm pretty sure both of those were inspired by mainframe operating systems. (I was a TRS-80 kid, and didn't know until much later why our equivalent of batch files had the extension "JCL".)

Having said that, though, we could at least wish that some variant of the Mac's old idea of separate creator and document codes stored as metadata had caught on. Sure, it'd have been a few more bytes per directory entry (to be specific, five more bytes!), but it was a lot more flexible -- and if more operating systems had been built with that "document types are metadata" idea, that metadata could have been replaced with MIME types later on, like it was on BeOS.

> But you're basically accusing programmers back in the late 70s of having insufficient foresight to see problems that would make their decisions seem bad almost half a decade later.

The URL standard was published in 1994.

Sure, in a vacuum you could read view my comment as meaning in all time, but given that the parent comment is discussing ICANN domains and my comment was relevant to that, I think it’s a little uncharitable to do so.

But it made them so much money!!

ICANN should never have done gTLDs at all.