Someone who doesn't understand what he's talking about telling everyone that we don't need X because the future is the cloud and the cloud is just so different.
If files could package and pass around metadata in the same way they pass around data, I might agree with Dare's conclusion. However, for me, metadata is becoming too important and has destroyed the utility of files (and pure data) being truly useful.
Example: Photos. The file is the photo. However, think of all the things besides that: face tags, text tags, album, ratings, etc. If you've ever tried to import 5GB of photos from iPhoto to Picasa (or vice-versa) you'll understand what I mean.
This is why "stuff in the cloud" is so useful: API support is there from day 1 so that you always know what CAN be exported and how.
"If files could package and pass around metadata in the same way they pass around data"
They can. It's all just bits. If your files don't and your API does, it is simply because your files don't and your API does, not because files can't and APIs can. Files can have metadata and APIs can be written that treat things as blobs.
There is a distinction, but it's not API vs. file, it's what semantic context you're in. Your MP3 player reads out the ID3 tags, because it's in a music player context. Your backing cloud store (like Dropbox) just sees blobs. Interoperability is less about agreeing on file formats than agreeing on semantic contexts. (Yes, file formats are important but they come later, and most arguments about them are actually arguments about semantic contexts.)
Saying "it's all bits" is too simplistic a view of file formats vs. APIs. I'm thrilled that MP3 files have ID3 tags, unfortunately that's an unacceptably small subset of the meta-data that I care about.
File formats evolve at a glacial pace and suffer from immense historical baggage. No one benefits from pretending that file formats are as accessible, agile or powerful as proper web APIs.
Strike the trendy "web" qualifier, and I'd argue we still have a problem, namely that "data" tends to outlive "behavior" by many, many years, and this fact is important and useful in both computing and civilization at large. This is why, in theory at least, relational data stores are more promising "trans-filesystem" candidates than either method-based object systems or "APIs". Hierarchy is inessential; data storage is not.
To say nothing of the fact that "APIs", Web or otherwise, aren't actually a means of storing data; at least in the context of things like "pictures" and "music", we obviously can never have "turtles all the way down."
Web APIs change on the whim of the provider and have no permanent basis to build on. No one benefits from pretending that web APIs are as stable, composable, or as powerful as proper files.
I italicized proper there because my real point is that you're simply defining yourself the win with the word "proper" and not comparing like to like. You can win any comparison by accounting for only the good of one proposition and only the bad of the other, but that's not a valid argument. If, for instance, we ask our web API to have some sort of assurance that it'll be there next year... hell, next week... suddenly "proper web API" turns out to resolve to absolutely no concrete noun at all. (How "proper" would you have called the Twitter API last week? You know, back just before Twitter started shutting huge swathes of third party apps out of it?)
But that's just one definition. The truth is you need the right tool for the job.
It is all, in the end, just bits. The bits don't have color [1], and they don't acquire color by being in a web API instead of a file.
I think that all of that can be supported by file formats. There is no rule that an image file format can only hold the pixel information. The issue is that it's hard for all vendors to agree on a standard, and once it's agreed upon, people want to add more information to it. We have EXIF for images, but what happens when iPhoto and Picasa both add face tags?
I don't believe files are the issue, it's agreeing on the interchange format that is.
It seems to me that any system or API you imagine or make up that associates a particular stream of bits with some sort of identifier is pretty much a file system.
Files will never go away. It all depends on the abstraction level that you view the term 'file'. Humans are files, a self contained unit of information (with built in copy semantics!) they have meta data (hair color, eye color, height, age etc). My point is that there will always be a container object of a sequence of bytes of data. Whether that's an mp3 file or a hummingbird, they are still files.
A more meaningful question might be to challenge the notion of a hierarchical file system. The web has managed fine without one, so possibly search across a flat collection is a better metaphor. There are numerous others.
There are, however, interesting possibilities that make files less conspicuous, apple are experimenting with this, where there are pipes between apps, I open a photo, send it to snapseed, pipe it on to tumblr. At no point am I confronted with a file save dialog. Obviously it's a file under the covers, but that's just an abstraction. Files themselves are abstractions. There's really no such thing as an mp3 file, it's just a sequence of bytes we can interpret as one.
Yup... another day on HN.