Comments on: Writeback, writing metadata back into your files

By: pvanhoof

pvanhoof — Thu, 12 Nov 2009 11:15:50 +0000

@Anders: the SPARQL update doesn’t need the tracker:writeback, the tracker:writeback goes into the ontology. We distinguish between user and miner by doing a bit of pseudo named graph support. If you pass FROM or INTO with your SPARQL Update query, we don’t trigger writeback. If you don’t and the property’s tracker:writeback is set to true, we do.

By: Anders Feder

Anders Feder — Thu, 12 Nov 2009 02:53:38 +0000

Looks very elegant. You must be avoiding a gazillion issues by having the tracker:writeback modifier in the update itself.

By: pvanhoof

pvanhoof — Wed, 11 Nov 2009 23:45:38 +0000

@Richard: you can have sidekick .xmp files with metadata written as XMP in, a lot of fileformats also allow embedding XMP of course. Like PDF.

By: Richard

Richard — Wed, 11 Nov 2009 19:41:35 +0000

My biggest qualm with metadata is that when it gets stored in a database detached from the files, moving files around and renaming them, unless I do it with a DB-approved application, or transferring them to another computer almost always loses it, unless I take care to locate and transfer the metadata database (which doesn’t work well when the destination computer has its own). I really do want metadata to follow the file around. :( I wish files in general had some standard method to store metadata that travelled with them, that they could exist in some file container that had [container header | metadata | actual file ] and that applications could deal with [actual file] or could deal with the container too. I’m sure that would baffle other OSes though :)

So, thanks for working on writeback support. The better attached metadata is to a file, the less fragile it is and the more portable it will be. YAY!

Now if I could only convince metadata systems to consider a hash of a file as well as its path to identify it.

By: pvanhoof

pvanhoof — Wed, 11 Nov 2009 19:03:21 +0000

@elLolo: When Nautilus uses SPARQL Update for that, yes.

By: elLolo

elLolo — Wed, 11 Nov 2009 18:51:02 +0000

So if I understand correctly it is about synchronizing data a user added to Tracker for a file (via Nautilus for instance) directly within the corresponding file?

By: fizze

fizze — Wed, 11 Nov 2009 18:30:21 +0000

Ingenious. It’s about time someone tackles this issue.
As this is in tracker, chances are it will be supported by a variety of platforms. Way to go!

By: Marius Gedminas

Marius Gedminas — Wed, 11 Nov 2009 18:25:53 +0000

This is very very interesting! (And a tiny bit scary.)

By: nona

nona — Wed, 11 Nov 2009 18:10:27 +0000

Brilliant.