OSNews

First off only two posts on d-d-l were about tracker solving a problem - its a bit exaggerated to call it spamming.

I can sure understand some people not wanting a desktop indexer but they are a minority and the majority of users (IE not geeks) will definitely need this technology.

Hopefully, when we are more integrated into the desktop, you will see the clear benefits (especially becuase we are not just an indexer but an RDF triple store metadata database too)

I haven't found the need to use desktop search yet (If you keep your files organized, you don't need to search), but I don't see a problem with the "tracker everywhere" proposal. *If* tracker is modular then several technologies inherent in an indexer should be applicable to many tasks. Why do we need to reinvent the square wheel? Doesn't Unix get it's enormous power by force-fitting everything to be like a file handle? If there's one central concept to understand, then implementing generic concepts that work on things that we haven't even thought of becomes a lot easier. It's called good design.

So here's my question to you. Do you see anything fundamental in the design of tracker that would prevent it from filling that generic role?

I haven't found the need to use desktop search yet (If you keep your files organized, you don't need to search)

Beagle is a huge benefit to me because of email and conversation searching. There is no real way to keep instant message conversations organized and when you deal with thousands of the mailing list emails it is very hard to find information within those emails no matter how organized you are. It also makes available many different types of information when you search which can be extremely beneficial. For example, I can do a search fo SELinux and find a paper on security stored as a PDF, email from security mailing lists, and news concerning SELinux from my RSS feeds. I never thought I would need desktop search either but after trying it out for a while I think it is the best technology to really mature in the past few years with Beagle, Spotlight, and Vista's new search.

Actually, there are a number of problems that Tracker solves (and solves nicely). For instance, open Rhythmbox with a library of 5,000 songs.

Wait for it... wait... hold on... Takes awhile to load, doesn't it?

Now, try importing some photos (let's say 100 RAW or large JPGs) into F-Spot. Watch that silly little progress bar creep along. You can keep reading along while F-Spot chugs, if you wish. Hell, you could order a pizza.

Now that you've finally got those two running (only two apps, mind you) check how much memory those hogs are consuming, just sitting there. On my computer, it's over 100 MiB. And I haven't even started using the damn apps, yet!

But if they simply made calls to Tracker when they needed information, they wouldn't have to hold a massive list of all the files, metadata, tags, versions, etc. in memory, would they (that's Tracker's job, remember)? Startup would be nearly instantaneous. And you wouldn't have to import your files into each application after you're already moved the files to your disk.

And, of course, search today is inconsistent across applications. Tracker is optimized for search (tag-based, keyword or service type), doing so better than either app, on more fields, yielding more relevant information. And Tracker can bring tagging to Rhythmbox and greater flexibility in sorting/organizing to F-Spot. For free! With less work! How nice of those Tracker folks.

With a single metadata store/indexer, the user gains consistency across the desktop, speed, memory and utility. Developers get to reduce the amount of application-specific code they need to write (think: far fewer bugs). Tracker promises to reduce development redundancy, speed up the apps you use every day and significantly shrink the memory footprint of the Gnome desktop.

Remember, my example only used two apps. Think of all the other projects on the Gnome desktop that are reinventing the "search", "index" and "tag" wheels. Tomboy alone sucks another 15-25 MiB of memory! And all it amounts to is a GUI for Tracker's tagging and object store capabilities.

So I've got to ask you, how is this bad, again? If I were you, I would read up on technology you're bashing before trying to discredit it. Your arguments are about as specious as claiming that filesystems or shared libraries are "solutions looking for a problem".