Big Fractal Tangle: all-hands interop

Everything needs to work with everything else. Can we at least agree on this? There's nothing worse than using a piece of software only to regret it later because it won't play nice with others. Even with things as simple as contacts and calendars, it's incredible how little we can interoperate.

How many of us keep our old contact manager around because it does certain things better than our new one, like printing labels or envelopes? Why can't we use both programs simultaneously without worrying about file formats or import/export? Just how many people start completely over rather than figure out how to transfer their data to the new program? Isn't it a bit amazing that we've grown used to this?

Not all application areas are so insular. Can you imagine if the graphics world had to contend with the same kind of vendor lock-in? All day, I open my PNGs, PSDs, and JPGs in a half-dozen apps, making changes with the program that suits my current need. Why can't I do the same with my address book, or calendar, or photo library? Why would *anyone* spend days and weeks tagging their photos with a program like Photoshop Album or Picasa or Jasc, knowing they'll be locked into those programs for the long haul, since there's no easy way use their data with other programs.

We decided months ago that we're gonna bend over backwards to interop with everyone. We want you to be able to open Tidepool to gather photos from your camera, open iPhoto to remove some red-eye and rotate a few, open Tidepool to do some tagging, open Photoshop Album to see those tags, etc. We'd rather compete on merit than hold our customers hostage.

Whether we'll be able to pull it off is another matter, especially given the proprietary vibe of most software companies. But we're prepared to do our level best to shame the rest into opening their formats or, even better, encouraging everyone to use open standards like we're doing. After all, what could be more semweb than all-hands interop?

Everything needs to work with everything else. Agreed?

Developing a standard surely helps. Thats what even XML standards like vCard, iCal or numerous others tried to do. Some of them have been ported into RDF, but just porting them into RDF doesn't help.

The applications today are tied to what they understand. The moment they get something else, they just don't work. What we need is a way where applications and systems can work even when they have a partial understanding of things they have at hand.

posted by Sunil Goyal at July 16, 2004 04:48 AM

In most software shops, adhering to standards is treated as a "nice-to-have", not a "must-have". Even if you're not intentionally polluting things (like Microsoft), it's all too easy to get lax and let just enough things slip to ruin people's day.

It's like reuse. Everyone talks the talk with regards to building reusable code libraries in house, but getting people to *use* them is another matter. Everyone's got their own damned way.

posted by Timothy Falconer at July 16, 2004 04:59 AM

Timothy, yep, this is definitely the direction in which we should be heading.

Leo - I was under the impression RDF/iCal was coming along quite nicely (although slowly - it seems like issues only get attention when they trip someone up). Timezones - as in W3CDTF, surely?
But whatever, the RDF approach does offer a lot more potential for reuse, as it's possible to map vocabulary equivalents after their creation. Ditto partial understanding - an unknown term my be a subclass of a known term.

Disclosure: I am a lousy reuser of libraries ;-)

posted by Danny at July 16, 2004 05:29 AM

a few principles we've bought into:

1) what doesn't bend will break

the success of the web is largely due to its looseness. As others have said, the 404 is both the best and worst thing about the web. Without this looseness, it probably wouldn't have taken off.

2) build them as islands, then stitch them together

heard timbl say this in nyc regarding ontology standardization, but it applies to standards in general. What's needed is the stitching part of course... a simple useful compelling way to say "this means the same thing as that" without having to get into defining delimiters and crap.

yeah, this is the way the semweb is going, but I want to see this kind of "esperanto mapper" in some software that's a little less complicated than protege.

posted by Timothy Falconer at July 16, 2004 05:39 AM