[Thunar-dev] Spatial or not-spatial?

[Biju Chacko]

Benedikt Meurer wrote:
> 
> Hm, ok, lets recall, for the WWW you use the query-interface because 
> thats the way to master a huge amount of data, but for your file system 
> you use the tree-interface because that's the only way to master a huge 
> amount of data - btw, to make sure nobody gets stuck on the terminology: 
> "data" = information, no matter if that information is presented by 
> local files, by static html page, by table rows in databases, etc.

I agree with you 100%. What is needed is desktop search. I have actually 
been thinking about the problem for a while.

There are two main ways of doing this: indexing or labeling.

Indexing is basically what google does: checks the content of files, 
indexes them and pops up a relevant file for a query. This does not work 
well for non-text data.

Labeling basically adds attributes both automatically and manually to 
all files. These attributes are indexed and used to generate query output.

The main problem with indexing is the number of kinds of files. In order 
to index every possible kind of file, you would need a parser for each 
and an ongoing indexing daemon.

The web, on the other hand, is mostly variations on html.

Labeling requires massive modification to everything in sight. 
Applications have to add attributes every time they write a file, 
file-save dialogs should allow additions of manual attributes and 
ideally all of this should be stored by the filesystem.

Indexing is easier to do, since it doesn't require buy-in from all and 
sundry. That seems to be the direction that others seem to be doing; 
beagle for example.

However, IMHO, the second seems to match the way people use computers 
better.

-- b