Two things I got way wrong, way back.

One was Linux on the Desktop (LOTD). Around the turn of the Millennium, I predicted big successes for LOTD and Linux on the Laptop (LOTL)—and continued to do the same, annually, until I gave up. Here's how my optimism looked in January 2003:

Is 2003 the year we see big-name hardware companies selling Linux as aggressively as they sell Windows—by which I mean Linux PCs show up in stores and on the front pages of Web sites?

Paul Saffo famously said we overestimate in the short term and underestimate in the long. But I'm going to go out on a limb and say the long term is mostly behind us. (How many years have we been waiting for LOTD to take off?) Somebody is going to break the ice this year. My own instinct says it'll be IBM, mostly because there must be serious demand for LOTD (as well as improvised solutions with it) inside the company. You can't have that many engineers walking around with Linuxified ThinkPads and not get around to selling them at some point.

I'm betting Dell will be next. Then Gateway. Then HP....

Didn't happen. Sure, some of the bigs sold personal Linux hardware (and still do), but not "as aggressively as they sell Windows". If you're not a Linux geek today, your choices of desktops and laptops are contained by two silo'd operating systems from just two companies: Apple and Microsoft. Hardware choices are also narrower since Microsoft decided to follow Apple into the hardware business.

The center of device use gravity has also moved over to mobile, where most of us are using gear running just two operating system platforms: Apple's iOS and Google's Android. True, Android is based on Linux (a victory we gladly claim), and the range of devices running Android is huge, but Android is still Google's show. The network frontier is also mostly fenced off by mobile carriers and device suppliers who believe, almost across the board, that captive customers and users are more valuable than free ones—and everybody agrees, at least tacitly, that "free market" means "your choice of captor".

Which brings me to the second thing I got wrong: expecting the world to prefer flat, distributed, open and free technology ecosystems over silo'd centralized, closed and proprietary ones. One example is XMPP, originally called Jabber. It was meant to bridge or replace all the competing proprietary instant-messaging systems in use at that time: AOL's AIM and ICQ, Microsoft's MSN, Yahoo's Messenger, Apple's whatever-it-was (now called iMessage) and so on.

Didn't happen. Look up "Jabber" today at LinuxJournal.com and you'll get a lot of results, also from back around the turn of the Millennium, mostly written by me. In those days, many of us had full confidence that Jabber/XMPP would do for instant messaging (aka chat) what SMTP/POP3/IMAP did for e-mail and HTTP/HTML and its successors did for publishing and all the other things one can do on the World Wide Web. We would have a nice flat, distributed and universal standard that people could employ any way they wanted, including on their own personal hardware and software, with countless interoperable systems and no natural barriers to moving data easily from any one system to any other.

Didn't happen. Yes, XMPP turned into a widely used standard and either supplied or inspired lots of messaging systems. But today, instant messaging is almost entirely silo'd, and it's more popular than ever. Facebook (including WhatsApp), Twitter, Google and LinkedIn have their own messaging systems inside their own "social" silos. So do Apple and Microsoft (including Skype). Tencent Holdings, the giant Web portal company in China, has Tencent QQ, WeChat and Qzone (a social site with messaging built in). Some interoperate, but most don't, since that would free their users from captivity.

Wikipedia currently lists 55 sites, services, protocols and programs that do chat. There is no underlying protocol for interoperation among them. Nor do any makers seem especially interested. Google Talk interoperated with other XMPP systems, but then it was replaced with Google Hangouts, which lacks support for XMPP. So, while XMPP.org says "it's not the end of XMPP for Google Talk", it hardly matters if Google Talk is itself dead.

The last nail in XMPP's coffin, perhaps, is this from Facebook:

On April 30, 2014, we announced the deprecation of the XMPP Chat API as part of the release of Platform API v2.0.

After April 30th, 2015 apps will no longer be able to access the service or API. This includes both access to chat.facebook.com and the xmpp_login permission.

We recommend people access Facebook Messages on the desktop via Facebook.com or Messenger.com.

On the other hand (or on another tentacle), the number and variety of chat platforms and apps is huge, and growing. Apple's App Store currently lists 171 chat apps, while Google's Play Store for Android lists 165. Nearly all of them are proprietary, and many also depend on (or work inside) large silo'd proprietary services. Listen for complaints about all this and you'll hear crickets.

APIs are also silos. As more and more of the connected world comes to depend on Whatever as a Service, "the cloud" and other remote and centralized services, the very idea of distributed capacities on standalone open-source (or open-source-based) hardware and software that individuals or organizations fully control as independent and sovereign entities in the world, seems terribly retro, utopian, or both.

Brian Behlendorf says, wisely, that the ideal is "minimum viable centralization". I think that should be a transcendent design principle. But the centralized thing is too often the easiest to make or to rely on (as is the case with APIs and big silo'd platforms). So now the chat ecosystem is a forest of silos standing on free and open-source geology, but not built with the same rock.

What may be hardest for old-skool types like me to admit is that silos are now the preferred method for inventing and promulgating technologies supporting uses that scale and invite incompatible competitors to do the same, while the marketplace as a whole doesn't have a big problem with any of it.

I'll never believe silos are the best way to make the world work in the long run. And I'll always believe that the flat distributed world built on free and open stuff is the most supportive and fertile base on which to build the best and broadest range of goods and services. But I'm past believing that this will ever be obvious to the world of developers, businesses and governing bodies. Muggles have always outnumbered wizards and always will.

Being out of sight and mind for most of the world doesn't make us wrong. In fact, it might just make us right in ways most will never see, no matter how much they depend on what we do.