Beyond Nautilus 1.0

[Heikki Keranen <heikrnen ees2 oulu fi>]

Hi

Here is some ideas what could be done after Nautilus 1.0 is stable.

Support for more file types

One of the problems in current Nautilus is that it supports only image,
pure text and audio file types. Audio files are not even visualized.
They could be visualized by standard waveform or spectral visualization.
In the future speech recognition could be used to produce text over the
waveform or spectral visualization.

Support for other file types like video files, word documents,
powerpoint files etc. should be created. Video files could be visualized
by the almost standard keyframe presentation (snapshots taken from the
video and presented as a comic like sequence). Powerpoint files could be
visualized very much a similar way. Word documents could be presented
like automatically derived table of contents (which could be reduced by
taking out subchapter titles to fit to the space required).

The question is: Is the basic zoom functionality in Nautilus 1.0
adequate? Many media types described above could benefit from the
"stretch" -option. E.g. with video file the zoom would affect the size
of each video frame and the stretch option would affect the amount of
frames in the visualization (zoom the time scale). Audio files has also
a time scale. Even powerpoint presentations can be thought to have a
time scale. 


Views

We could have more media specific views (like the "view as Music" view).
Window could be seen as a 2-dimensional RGB or HS(V) color space and the
images could be plotted to the window by their average color. If the
user is looking for a red picture he/she could scroll the the window to
the direction where background is more red.

There could be a time line view where the files are plotted by their
creation, modification or whatever dates. When the location services
becomes available and the place where the picture is taken with a
digital camera can be stored in the image automatically, it would be
handy to have a geographical map view, where the thumbnails of the
images are plotted. OK, this will not probably happen very soon, but I
think this is going to happen some day (when there is a GPS-chip in
every digital camera).


Search functionality

Pre-search indexing is required. There is no other way to make searches
fast enough. It should be 0.1 seconds instead 30 seconds to scan your
entire hard disk. In addition, we need to index not only normal file
attributes (file name, modified date, etc.) but also content of files.
In image files this could mean color distributions (histograms). Sound
tracks in videos could be driven trough speech recognition software and
indexed. This way the user could easily find all video clips from
football games were the speaker shouts "GOOOOOOOAAAL!". At least in a
theory.

If search speed is high enough (10 queries/second), queries could be
made dynamically. This means that results are updated "on the fly" while
user modifies the query.


How could this all be implemented? (for programmers only)


Short answer: Good modular object oriented design. 

Longer answer: To have a sufficient search speed we would need some kind
of database to store the indexes (MySQL or whatever). However the
problem with standard relational databases is that they don't do very
well with multimedia indexing which requires multidimensional indexes.
But I think it is possible to find a solution for this.

When the users want to access network disks, the server computer could
perform the search. The problem is that the current networking protocols
don't support this kind of searches. The complete solution is to
update/recreate those protocols. We should take into account also the 
wireless devices. In the short term XML files could be used to store the
the indexes. Creating those indexes is the most time consuming task and
should be done in the background (a way to utilize the spare processor
time).

It should be easy to expand the system by plug-in modules which
understands certain media types. In the long run the application (and
file format) designer and programmer should also make the plug-ins 
which can understand, index and search the new file format. This search
platform could be utilized also in "inside application searches" (Edit
-> Find...).


I think this kind of system could help users to find files (and file
contents) much more faster. Think what an enormous amount of time is
wasted every day to find the right file or a piece of the file). This
could be also a way to get ahead Microsoft.

I'am doing my master thesis about how to find a right piece of media
with a wireless mobile device and I have some ideas how to do these kind
of things. Unfortunately I have hardly any experience in linux
programming (except Java) and not very much time to learn these things
but I could offer help and advice if anyone is interested to
explore/implement these ideas.

Comments?

	Heikki Keranen