There’s been quite a lot of buzz around so-called ‘search’ tools lately, from the on/off WinFS, to Beagle and Apple’s Spotlight. Tenor is KDE’s proposed ‘framework’ for taking this even further. Kurt Pfeifle, along with insights from Scott Wheeler (one of Tenor’s authors and primary designers), sat down to talk about the current problems, Tenor itself, how the thinking is different and how KDE’s flexible technology provides a solid basis for making it happen.
As Im rerading it says to work the application must use the Tenor engine to be able to work with, from a point of view of a KDE application that would be nice bur fro a none KDE application would be a limitant.
The tenor backend will be independent and any application can tap into the Tenor database if configured to to so.
Tenor is going to rock. It is so much more meaningful to look for relationships among your files than to simply search their contents, which Tenor will also do.
The attached articles do not fully convey how revolutionary it is. A better link might be this one:
http://dot.kde.org/1109163846/
It strikes me that KDE is in the almost unique position of having the technology, the development ethic and the organizational structure to accomplish such a turn-around, thanks to the tightness and code-reuse within the KDE platform. Which other player in the field could roll out such heavy-weight technology in a reasonably complete set of desktop applications at the same time – the enabling precondition for being able to take advantage of contexts? Apple comes to mind, but the list pretty much ends there.
It’s pretty lovely you can implement much of the necessary data mining directly in the KIO slaves, not even touching the app sources.
Storing metadata about the relationships between files and not discarding this important data when doing daily tasks like reading mail and surfing the web is not a new idea. I remember discussing this on IRC a couple years ago when hashing out dashboard ideas. It’s a problem that keeps coming up again and again. It’s great to see developers starting to really tackle it. The beagle approach seems to be a two-phased attack on the problem. Phase I is to get the problem of searching contents of files and indexing that data and allowing for searching of that data. This is important and you can do that in Beagle right now. Phase II involves storing the relationships between things that you do on your computer. Read more about it here:
http://beaglewiki.org/index.php/Beagle%20Roadmap
I think Tenor looks promising and I hope the developers of Beagle and Tenor can get ideas from each other. We’re starting to see a lot of innovation happening in KDE and GNOME and I think that’s awesome.
It was eye-opening to see the sort of file counts the author deals with on a daily basis. Every time I see an article on a new file search engine, I’m hard-pressed to see where I would want it for myself. I guess I’ve trained myself to use hierarchies well and I reorganize them aggressively as the need arises. Most importantly, I rarely have trouble finding files I want to look at. The difference is that I have much fewer files than I guess most people have, and I usually deal with them in groups, e.g. TV show, programming project, etc.
I’m just starting to get into digital photography, and there’s no way I’m going to bother with traditional hierarchical organization there. I can see this being a real problem in the future. It would be cool if Tenor, Beagle or something like that could help me search pictures. I’ve also wondered if anyone has worked on visual search methods for images; I’m intrigued by the possibility but don’t know a thing about the research in the area, if there even is any.
I have nothing against these new trends for storing/searching info in computers. But traditional folder based storage seems enough to me. It is just a mather of how organized every person is and how much computers should need to deal with organization for us . As I said, I have nothing against people using these technologies as long as they are not forced to be used for everybody. I use gnome, for instance, but I won’t be using Beagle, I just hope these things stay as alternative options you could decide to use or not.
“I’m just starting to get into digital photography, and there’s no way I’m going to bother with traditional hierarchical organization there. I can see this being a real problem in the future.”
Actually most photographers are rather strict control freaks. Its not even a secret outside of the industry.
When it comes to large jobs, and having to keep track of multiple jobs being processed at once – its hard to say anything is more convenient than using a hierarchal system of directories and file names that include job numbers, dates and version sets. Especially considering the fact that its possible to set this kind of a filing system, with a simultaneous archival backup, to run automatically from with in a basic pre-post-production Photoshop macro for each session import.
Photography isn’t exactly a bleeding-edge, incorporate the newest wiz-bang buzz-tech into my workflow kind of field — especially when it comes to managing your assets.
I don’t think it should replace the hierarchy, just supplement or provide additional functionality. I use Google desktop from time to time at work. But it certainly won’t replace storing files in a hierarchy.
Yes, this looks like a great idea. I’ve thought about this for a few months now. Keeping track of contexts and file relationships could be revolutionary to desktop search. This clearly seems to be tackling the real problem that Spotlight and Google desktop don’t which is: how to create meaningful and useful meta-data for files. It easy to do stuff once you have meta-data, but getting it is the trick. It has to be painless to the user, or it won’t be used.
This won’t replace file heirarchies, it will just be another tool to help us find and keep track of files, and their relations.
It would be really nice if linux distro’s started adding more meta-data to all the standard files, especailly things like what package it belongs to. It could make dependancies much easier to settle, especailly for developers and packagers. It could make package tools work flawlessly. Another example is for build scripts to put their mark on files they install, you can alsways know what scipt installed what file, and how to easily get rid of a program.
Whatever is the result out of the Tenor/Beagle interaction, make it a FreeDesktop project. Can’t have two different desktops using different mechanisms to make semantic desktop.
Let me give you an example of where this sort of thing becomes useful. In the process of researching a project I’m working on, I’ve collected hundreds of pages of documents, ranging from PDFs to text notes (taken from microfiche archives), and dating from 1964 to 2003. Needless to say, there is absolutely no existing order to these documents. Creating a rigorous ordering system for these things is an enormous pain, and the standard filesystem hierarchy is absolutely useless for defining such an ordering. I’ve resorted to manually maintaining a giant text file containing all the metadata. Something like Tenor has the potential to be very useful in a circumstance like mine.
Of course, my example is actually a rather trivial one. I interned at a company once that needed to manage all the acquired knowledge of dozens of different NGOs (eg: The Red Cross) from locations all around the world. Basically, if some doctor working for the Red Cross in Ghana wrote a document on “best practices” for treating malaria, a healthcare director working in DC needed to be able to access it, and pass it along to his workers in Bangladesh. There is some very specialized (and rather expensive) software for dealing with this problem. While I doubt Tenor will be able to handle problems of the same magnitude, there is a good possibility it will be able to handle problems that are similar in nature, though perhaps (much?) smaller in scope.
I think the answer to your question lies in the technologies that each desktop has, KDE has some great file handling capabilities already built in and it would be a shame not to use them and Beagle is based on Mono which I don’t see getting near KDE with a ten foot pole. It would be neat if they shared ideas, pitfalls, and such.
come on….. this seems to be the new thing but i dont see it working very well, we currently have folders, and long filenames, if that isnt enough organization for you then i dont think anyting else is going to help you
“searching” your computer system will be the same as doing a search on google – millions of hits to look thru and still hard to find what you was looking for….
if i cant keep track of what i want using folders and long filenames then i obviously have too much stuff..
The searching and keeping track of stuff in the old sense of thinking are not really the thing with Tenor. Using well organized files or google like searches are parts of what it can handle, but to me the main exiting points are the things you can get out of the information ‘process’ with Tenor.
Things like tracking the information, showing how the picture in your wordprocessing document is from a mail person x sent as opposed to trust you memory or having written the origions of it somewhere. Giving you enough information to be able to ask person x for the next picture in the series, as you need this for you next document.
Or the possibilities to tie information together without moving it around. Sorting in folders and filenames have it’s limitations. Like a file related to two different subjects, in the old way you could duplicate it or perhas on the filesystem level use symlinks. Both methods have weaknesses and don’t scale well when introducing things not filesystembound, like email.
That’s the problem. Developers are lazy by nature, that’s why they write programs to do things for them. The reason DCOP worked so well was that it didn’t require any action from the developer, any actions in the program would automatically become DCOP actions. For this to work it needs to require zero effort to implement, and not get in the way.
“I’ve also wondered if anyone has worked on visual search methods for images”
You could have a look at ImgSeek (http://imgseek.python-hosting.com/). It’s a python/qt image library with an interesting search feature: you can search an image by its similarity to another, or even to a quickly-drawn sketch.
I’ve heard Intel research was working on a such a library, but I dont’ know if anything became of it. Anybody know?
You could have a look at ImgSeek (http://imgseek.python-hosting.com/). It’s a python/qt image library with an interesting search feature: you can search an image by its similarity to another, or even to a quickly-drawn sketch.
Unfortunatly it’s too color biased. Even if you search a high-contrast image, a near perfect sketch of the image searched will be next to useless because the first 20 hits look absolutely nothing like the image -or your sketch- but have a spot of exactly the same color you used somewhere in your sketch (not in the same spot or in the same shape, just the same color). If you have a good memory for colors imgSeek might help you otherwise tenor or traditional methods are most likely better.