getwired.com

As I sit here writing up my final post on media metadata, it hits me… one of my least favorite tasks of writing a new blog post. Tagging.

Why does search have to be such an afterthought? Why is it that the operating system has to have such deep insight into binary file types? Because nobody ever had the foresight to say that people would want to index everything – and that only odd content would be unindexed.

I had lunch at Terra Burger a few months ago not long before I started writing this series of posts. I bumped in to a parent who, like I, was trying to get used to using his new Mac to edit video and photos. He had been using the Faces component of iPhoto, and was disappointed at how it confused photos of his 1 year old son with photos of his 4 year old son that were 3 years old. Funny, isn’t it – how “search” indexes can be thrown off by just failing to take one metric (the age of the photo) into account? The other glitch was, as I have recently found, that iPhoto is handy, but that Faces – like all other postscript indexing (indexing after content generation has been completed) is a pain in the butt and consumes time that I sure as heck don’t have.

I think that the future lies in devices that are more indexing aware – GPS-savvy cameras, camera phones (the iPhone did a great job of schooling the industry here), video cameras, and more. Content itself will grow to be more natively indexable. MIT Technology Review had an interesting – but fundamentally flawed – article a few months ago discussing “open video”. While I will be discussing video technology more in the future (given my new job), it’s important to bring up this article because I believe that while the technology just doesn’t work the way the author defined here, that it does highlight something that does need to happen. That is, that video content needs to become innately searchable. Not just by Google, but on your local computer, on Facebook, everywhere. The fact that content is in a binary compressed streamed form (making it inherently hard to decompile) the content itself should instead provide for a first-tier indexing experience by instead promoting it’s own manifest of what the content is. Inherent content indexes (name, date, generator) explicitly defined indexes (creator, location, participants) and content (objects, places, definitions, categories and tagging, a storyline/scene flow, and of course a good transcript of any spoken word – often relatively easily done via speech recognition – albeit with the potential for flaws). Just as interesting is the relationship of this snippet of content with any other pieces of content created in the same medium, at the same time, etc. The genetics of content are just as important. Knowing that the tiny 250KB JPG came named f34af3.jpg came from the original image D642242.jpeg offloaded from your camera on 12/4/2005 – all of that is useful information. Just as crucial? Finding duplicates in a useful way and sharing that data across content consumers (spouses, grandparents, aunts and uncles, etc).

Think about it – today, search indexers must go out of their way to juice those pieces of data out of a chunk of content. If all content exposed its metadata in a uniform, easily consumed way, anyone could index it – Google, Spotlight, Microsoft Desktop Search… anyone.

Indeed the future of content search and metadata success lies not upon better search – but rather in better metadata exposure, and more of that being populated and published as automatically as possible.

People want to find things. People don’t want to make them searchable.

On Twitter, I was reminded of course that metadata (the lack of it) isn’t just a home media problem. It’s pervasive in our lives – especially the more you let technology into your life. I’ll expound upon that later.

In my first post I mentioned WinFS, and why it was symptomatic of the “metadata problem” that we all live with today. I’ve chosen to hone in on home media just because it’s something that we all live with – specifically the problem I mentioned earlier, where we all have media goo that we’ll never share again. Those memories that you took the time to photograph or record – may as well be buried in a cave somewhere never to be seen again.

The key problem here is two-fold. 1) You’ve recorded onto “analog” media. Hey – even if it’s a DVD, you have no way to truly “search” it. Photos are a “hand index” media only unless you begin with digital photos (check out the upcoming post on iPhoto and iMovie as they relate to that). 2) Any references that you may have had to the content of the images/video become lossier the longer you go from the time of capture to the time you try to “catalog” them. You can’t remember which day was which, which cousin was who, or where that boat tour was, and what the name of the lake was that you went across.

Truth be told, we’re all innately horrible at capturing these kinds of details about events and memories. Only the lucky person gets to recall exactly how to get back to where they were driven once without needing a map or directions. Most of us need notes, maps, or other tools to recall the small details – the kinds of things you want to recall when viewing the photos or videos with the kids a year later.

When was the last time you set the metadata properties for a Microsoft Office document you were working on? Wait – you didn’t KNOW you could add metadata properties to Office documents? Well – even if you did, you haven’t set one more times than the number of thumbs you have. I know. Don’t lie to me.

For this reason, I am electing to define two types of metadata. Intrinsic – that which can be innately, directly gathered from the media itself, and extrinsic. My example in my first blog entry in this series, the above example of Office documents, and to a large degree WinFS’ design (as most of us would have experienced it) are all extrinsic. Much like taking the time to catalog a series of 35mm photos or slides, or edit a bunch of VHS-captured memories into any form of tolerable viewing (perhaps even with captions or cataloging), nobody does this. We don’t have the time to do this – at least more than a few times and then we tire of it. Thus, “memory to media goo”. The cool new device or media type becomes frustrating because our initial intention – to share memories with others or preserve them in a useful way for ourselves, is just too damned hard.

Simply put, using extrinsic metadata to organize anything sucks. Even if it works in theory, it doesn’t work at scale, in real life. We all give up and stop trying to use it for all but special cases.

Instead, intrinsic metadata is the future. In my next post, I’ll be discussing intrinsic metadata, what it is and how it works (when it does) and where we’re all going from here.

Five years ago this week, I left Microsoft. After more than 7 years there, it wasn’t an easy decision to make – but I did. This was after almost 5 years working on the Windows team. During the spring of 2004, when things weren’t honestly the rosiest for Windows Longhorn (what would, after a code reset in summer 2004 followed by much cutting and pasting, become Windows Vista), a co-worker in Windows Marketing and I were having a conversation. It went something like this:

Him: “Have I shown you the beautiful pictures I took on my trip to Egypt with my wife?”
Me: “No – show me. WOW, those are amazing.”
Him: “Thanks – was a great trip. See…” (clicks a button to winnow down the visible set of photos) “these are the ones at Giza…”
Me: “Wait – how did it know that?”
Him: “Oh, I entered the information on each one as a keyword”
Me: “So… you entered keyword info on each of the photos you took?”
Him: “Yup. All of them.”
Me: “How many were there”
Him: “Several hundred”
Me: “You don’t have kids, do you?”
Him: (grins)

This was especially amusing because he actually was the Product Manager (read: the guy who owns the “marketing story”) for WinFS. For those not familiar, WinFS was the abandoned strategy within Microsoft (Windows Longhorn in particular) that we had first announced at PDC 2001. WinFS was an attempt to inject “metadata into the filesystem” – or so the world was told. I won’t go into my entire WinFS tirade here… Grab me sometime and I’ll tell you a story over a beer.

WinFS’ technical foibles aside, it suffered from a classic software problem – solution naivete. Think about it. The story above… How many of you have dust-gathering stacks of:

1. Decades of un-annotated, uncategorized 3×5 photo prints
2. Negatives (or CD originals, now) of said un-annotated, uncategorized photos
3. DVD/CD-R/8mm/Betamax/VHS (or other) video content of barbecues, family reunions, birthdays, bar-mitzvahs, etc
4. Reels of even older audio or video content
5. Slides (yes, I said it, SLIDES!) – my dad fell for this one

Ready? We all do. Yes. Admitting it is the first step. We all suffer from what I like to refer to as “memory to media goo”. It’s where you transact your memories to media, instead of your brain. Never to be seen again until… you finally clean out the closet (as we did several months ago).

The only problem? By the time you find this media again, you can’t remember if that’s uncle Phil, or uncle Bill. You can’t remember if it was your brother’s wedding, or your sister-in-law’s. No, instead, you’ve now forced yourself and your kids to go through ALL the photos again. Even the ones with red-eye, since those came back from the K-Mart photo processing lab too, and the entire 5 hours of VHS from little Timmy’s first  month of life. Not because they’re interesting, heavens no. But because those memories are glued on that media. For good. And there is no way you can pry off the “interesting bits” off and throw out the other 95%.

In my next several posts I’ll discuss exactly why we’re in this rut, how things are getting better (how you can make them better for yourself), and what this means for consumer media and software (hint: it means we’re all headed for a few compromises, where the best format may not win, just like the old Beta vs. VHS war).

SiriusXM’s stock is officially on the floor. And it’s had me thinking why – as I’ve been pondering canceling my own subscription.

Since Howard Stern went to Sirius in 2006, he has been talking big about the death of “terrestrial radio”. But I think it’s bigger than that. With iTunes, a growing selection of on-demand media (think of all of the places you can on-demand movies from), and to a smaller extent, the growing reliability of Internet-based music and media channels, the threat posed to both traditional radio and SiriusXM is real. Add to that the shrinking volume of disposable consumer dollars, and SiriusXM is as hosed as “terrestrial media”.

Honestly, the music selection of SiriusXM isn’t that spectacular – the playlists repeat more often than they should, and the variety means that personally, I have about maybe 15 channels, tops, that I ever listen to. The sole breadwinner that they have is truly unique broadcasters – such as Howard, Oprah, and Martha… I listen to Howard for the whole rubbernecking factor – just to see what happens next (and frankly I’m ashamed of myself ).

But it’s getting harder and harder for me to validate paying for SIriusXM – when Howard and a few other channels are all I would miss – and I have to think, post-acquisition, that many Sirius, and LOTS of XM subscribers, feel the same.