"The Future Of Spotlight"
Hi folks,
Just came across this article which I found via osnews.com: http://slashdot.org/comments.pl?sid=146369&cid=12266911
Some nice Ideas about possibly upcoming new Spotlight features in there.
Have a nice weekend, everyone!
durandal
Just came across this article which I found via osnews.com: http://slashdot.org/comments.pl?sid=146369&cid=12266911
Some nice Ideas about possibly upcoming new Spotlight features in there.
Have a nice weekend, everyone!
durandal
Comments
Originally posted by durandal
Hi folks,
Just came across this article which I found via osnews.com: http://slashdot.org/comments.pl?sid=146369&cid=12266911
Some nice Ideas about possibly upcoming new Spotlight features in there.
Have a nice weekend, everyone!
durandal
You forgot to mention, that the writer of that post seems to be an insider. Interesting stuff pointing to more generated metadata og builtin GPS in powerbooks.
We're talking about highly advanced stuff here. It exists only in labs. So it's way too early to talk about specifics.
I don't want to blow anything out of proportion, but think of Spotlight as being kind of like the first bitmapped graphics. What we're doing with it right now is cool. But what's really important is what it enables us to do in the future.
GPS-based locational metadata is just one example. Automatic speech-to-text transcription for audio recordings is another. (You wouldn't believe what vector processing can do for speech-to-text. I saw a demo where a high-quality, noiseless audio recording of an unaccented speaker was transcribed at 20x real-time on a single 2.0 GHz G5.)
Example: You're doing a multi-party teleconference. A recording is made of that teleconference (each angle), and separate audio tracks are recorded for each participant. In real time, your computer transcribes each voice track and stores it as ancillary content on the recording, content that Spotlight indexes for you. At any time, you can type "meeting in San Jose" into Spotlight, and it'll take you right to the angle and track on which your co-worker Laurent talked about next week's meeting in San Jose.
Think about more detailed logging. Right now your computer logs only the most rudimentary events, stuff that is of no interest to human beings. What if it could log everything? Right now you can say "Show me that file I worked on yesterday at two o'clock." But what if you could turn that around and say, "When and for how long did I work on this file?" That's vitally important to anybody who does billable work. Imagine if, through metadata and logging, your computer could automatically produce your time sheet for you?
Another type of automatically generated metadata we're experimenting with is relational metadata. Let's say you've got a picture of your dog on your computer. You e-mail it to your sister Jan. Your computer notes this as metadata on the photo so later you can ask your computer to show you what pictures you've sent to Jan.
Address Book is one area where relational metadata is pretty important. In Address Book, you put Jan and your brother Harry into a group called "Family." Both Jan and Harry, in their contact records, get metadata describing them as being members of the "Family" group. So later you can ask your computer to show you what pictures you've e-mailed to members of your family. Or received from members of your family. Or what pictures you've e-mailed to SOME members of your family but not ALL.
Let's say you take that picture of your dog and drop it in a Pages document, then export the document as a PDF and mail it to your sister Jan. The computer records, as metadata, the fact that that picture of your dog is related to Jan. It knows that put associated the picture with that Pages document, that the Pages document was associated with the PDF file, and that the PDF file was associated with an e-mail to Jan.
Now combine it with a gestural interface. Take two files, any two files. Say it's a PDF representing an invoice and a Photoshop file representing a poster you designed. You drag the invoice over the Photoshop file and a marking menu appears, giving you the option of establishing a relationship between the two files. If you want you can annotate the relationship. If you don't, you don't have to. The computer will simply note that a relationship exists.
Now extend that idea. Instead of it being two files, it can be two ANYTHING. Drag a contact from Address Book to a Pages document; up pops a marking menu asking you if you want to establish a relationship. Or an song from iTunes to a picture of your girlfriend. Or your daughter's birth certificate to her birthday in iCal.
The possibilities that Spotlight opens up are pretty inspiring. It's not just a desktop search tool. Yes, it makes that possible, but bleah. That's 20th-century thinking. That's you working in the way the computer wants. What's more important about Spotlight is the fact that it's an enabling technology that lets the computer work in the way you want.
There's some pretty exciting stuff coming in the next few years.
That way, when you upgrade and transfer your personal folder, all that info will get saved, and the system wide stuff can just be regenerated.
Originally posted by WhiteRabbit
Yeah, cause I mean the guy's totally right about, personal actions data becomming important in the future. Although I don't think that it should become part of the metadata of the file, it should be linked in some way. Because you don't want the file that you sent your friend to end up with all that data about you. However each file your computer could have a link to a a file in which is recorded all actions that that user does which relate to that file. When you send the file to someone else, it will record that action in the action-history file but since that isn't acutally part of the file, your friends won't get that info. His computer will create a new action-hisory file in which the first entry will be that the file was sent to him by you.
Yeah it can't be on the actual file for obvious reasons. Umm, maybe Apple have a trick up their sleeves - .Mac!
All that really excites me !!
mm.. may be i should apply for a job at Apple.. so, may be i should finish my physics graduation after all ...
Originally posted by MacCrazy
Yeah it can't be on the actual file for obvious reasons. Umm, maybe Apple have a trick up their sleeves - .Mac!
My GUESS (I don't know for sure) is that Spotlight uses CoreData which, in turn, uses SQLite database API for storing information. When you first install Tiger (as per the reports), you have to do an initial indexing of the file system for Spotlight, and my guess is that it takes any and all metadata it can derive from all files and stores them into its SQLite database.
You can read more about Coredata and SQLite here:
Developing with CoreData
Originally posted by atomicham
My GUESS (I don't know for sure) is that Spotlight uses CoreData which, in turn, uses SQLite database API for storing information. When you first install Tiger (as per the reports), you have to do an initial indexing of the file system for Spotlight, and my guess is that it takes any and all metadata it can derive from all files and stores them into its SQLite database.
You can read more about Coredata and SQLite here:
Developing with CoreData
My guess is that you're very wrong. SQLite great for portability, but is not really great performancewise. IIRC SQLite doesn't support typing, which i recon must hurt performance. The spotlight index isn't even SQL.
My guess is that the whole datastructure and querying in spotligt is completely custom built.
Originally posted by PrettyBoyClone
My guess is that you're very wrong.
Certainly wouldn't be the first time...
My guess is that the whole datastructure and querying in spotligt is completely custom built.
Whichever the case, the point goes back to WhiteRabbit's concern that this data would be stored in each file. We agree that they are storing the database of info separately from the files themselves. There isn't concern that a random text file (say a .c) is going to contain a spotlight reference to some personal piece of information.
Originally posted by atomicham
Certainly wouldn't be the first time...
Whichever the case, the point goes back to WhiteRabbit's concern that this data would be stored in each file. We agree that they are storing the database of info separately from the files themselves. There isn't concern that a random text file (say a .c) is going to contain a spotlight reference to some personal piece of information.
As i understand it, the metadata is stored in the actual file and the index (and metadata relations) are stored on each volume.
-So when you move a file to a new volume, it is indexed there.