Social Semantic Sense for the Desktop (MIT Technology Review)

PS - in this case, the full sentence being nitpicked is in fact "People naturally group information by topic and remember relationships between important things, like a person and the company where she works." The example given is of two things, and it seems to me that the relationships being considered are in general binary relationships. (I can't think, off-hand, of non-binary relationships that cannot be decomposed into binary relationships.) So both premises of the nitpick were wrong.

So being an old fart doesn't make you a smart, or even a considerate fart. :-)

I guess we all agree that the shoe that fits one person doesn't necessarily fit any others' person's feet, and that shouldn't be discussed. Nevertheless, I personally found your comments below the relatively high level of both, intelligence and respect I'm enjoying reading LWN for.

Nobody forces anyone to use a desktop indexing system (which is what you seem to be talking about). I somewhat see your point, until now, desktop search systems I've tried didn't cut it for me. (And I didn't even consider running it on 15-year-old hardware like you seem to be enjoying.) Mainly because of two points: resource consumption and lack of integration. (Basically they're too slow and too far away from what I've been doing.)

The point of the NEPOMUK technology though goes much, much further than indexing of files on the desktop. It's about semantic technology that can be used to organise users' data in a way that comes closer to how the human minds organises the data.

I see NEPOMUK very much like the next big thing, the enabler that sets us apart from proprietary technology. Implementing the semantic desktop, a desktop that's smarter and presents all the information you store on it in a way that's understandable and recognizable for human beings is one of the next big things that we're planning to do in KDE. And I think we have the foundations in the technology at our fingertips now. So let's have a look at how this could look like, and some use cases along with it.

We're started off with desktop indexing ("find me files containing 'fnuiken'"). We've added manual tagging and rating of files on top of that ("Show me all files belonging to project XyZ"). This is where we are now, pretty much, although the lack of user interfaces available for that make it all a bit hard to discover.

Next up is making applications provide meta information automatically (think of storing information of origin, usage and context ("You've downloaded this file from foobar.org, you last opened it when viewing holiday photos on 13 December"). Integrating information from ontologies would be a following step ("This photo shows your friend Max, click here to chat with him" as another cheesy example).

We're only just touching the surface of what a technology like NEPOMUK enables us to do. Dismissing it as waste of resource, while fine on a personal level, doesn't really do justice to its potential, despite its horrible name ;-)

"Probably considerably older than you if push came to shove and have been involved in the PC side of computing almost since day ONE ."

Ooooh, I'm impressed. Or maybe not.

"and yes microSloPs produces windBloWs"

Like I said: 12 year old...

"also if you have a decent file system layout you know where your files are stored "

Manually locating files or folders (eiter with a filemanager or through CLI) is A LOT slower than typing few letters in a search-tool. Even if you have the worlds smartest filesystem-layout, you are still a lot slower at typing/mousing than your computer is at searching.

So try a better argument.

We are fast approaching (or we have reached it already) a point where you could have a whole core in your computer dedicated at running various housekeeping-tools (indexers, spellcheckers etc.) and it would have a neglible impact on overall performance.

What you want to do is to do everything yourself. But you fail to understand the fact that CPU-time is cheap. Very cheap. But the users time is NOT cheap.

>Linux systems come with the find utils and the locate DB that is plenty quick enough unless of course your are a speed freak also if you have a decent file system layout you know where your files are stored

I've never liked locate because it thrashes the disk every single time it updates the DB (destroying the disk cache, as well as making most all desktop applications slow as hell). Locate also sucks w.r.t. staying up to date (it doesn't, for only finding system files this is okay but sucks when trying to find your own data). rlocate avoids always being out of date, but doesn't solve the disk thrashing or only operating on filenames, as well as requiring you to run a daemon and kernel module.

Having a decent file system layout only helps when you have a small number of specific files. As you get more and more files, a hierarchy approaches uselessness for *fast* retrieval, often simply getting in the way.

Desktop search isn't meant to replace locate or filesystem layouts, it's meant to solve a COMPLETELY different issue, one that deals with an explosion of information that requires the entire system to cooperate to successfully alleviate.

IMO, NEPOMUK is the _very_ beginning of the road to a desktop search that doesn't suck. The ideas behind nepomuk is the answer, it'd be great if more work could be put into it (especially app support).

>i find them an absolute total waste of time
Right back at ya. They're inadequate for mine and many other peoples' needs. It's high time we advanced beyond hierarchical representations of information that are merely a vestige of the limitations of an analog world.

If you're this petegn:

http://www.bebo.com/Profile.jsp?MemberId=1562301261

then you're full of it.