I do it like this - import the photos to a folder, go trough them in a fullscreen GQView and delete all the blurry ones and redundant copies (leave the sharpest one) and after I have a set of worthwhile photos I import them into F-Spot for tagging and further management.

I think I'm probably getting unduly hung up on where the meta-data is stored.

So far it seems that most of the managers I've looked at store all the tag details within their own database - which over time will provide lock-in.

For the moment I'm doing a similar thing to you, I'm taking pictures, moving them into a directory tree then storing the meta-data inside the files themslves.

Forexample:

exiftool -comment="Tag:indoors Tag:foo Tag:bar" img_1.jpg
exiftool -title="This is the caption" img_2.jpg

That seems sane to me, but I'm going to have to look around some more, because it does feel like reinventing the wheel.

When I first got a digital camera I organized photos this way. Unfortunately it started breaking down after 10,000 photos and I ended up going back and sorting everything by date, with each shoot in its own dated and named directory(IE, all photos from an Aurora shoot in a directory named 2006_10-06_KenaiBeachAurora) and then each years shoots in a directory for that year. So it would be skx@gold:~$ ls /home/skx/archives/2006/2006_10_06_KenaiBeachAurora

This gives you a file structure for sorting photos, and removes alot of lockin you see.

My search for my first photo organizing program lead me to digikam . Metadata was at the top of my priority list, and I found digikam good about (optionally) keeping it in the files themselves, and being thorough in its expression (jfif/exif/icmp/xmpp). I also appreciate it being directory structure agnostic (if you want).

When I started photographing, I was using iPhoto on Mac OS X, but when I switched to Linux I first used digiKam and then F-Spot. They're both pretty good, but F-Spot's GTK goes better with my GNOME desktop than digiKam, and I appreciate F-Spot's concept of 'multiple versions' of a photo, so that if I open a photo in GIMP and later decide I messed up the colour balance or the cropping, I can always go back to the original.

F-Spot and digiKam both have their own databases, but both also have the option of copying tags and other-metadata into the EXIF chunk as well - when I imported my photos from digiKam to F-Spot, I just said 'import from this directory' and all my digiKam tags suddenly showed up in the F-Spot tag list.

Fspot has an option to store the metadata in the exif data
as well as in it's database.

I've been using digital photos for 10 years. My prefer now is to keep "Master" images in YYYY/MM/NN-DESC. Where NN is 01 to 99 and DESC is a short description.

I find that image management packages like workstations come and go, but if there is a picture I want to find I can generally remember the date I took it.

The DESC is also find-able. If you rename the photo files with more detail they are also fine-able. You could also add find-able tags by creating empty files with the appropriate names in the directories.

Image packages are useful if you want to do some batch editing or bulk uploads. A good file system layout with last a long time.

You can then layer another system on top, but if you lose that system then you would lose the basic structure.

Thanks for the comments everybody - I think that I've now got my mind made up:

• I'll archive images to sensible local paths, probably date-based rather than my current description-based system.
• I'll store any details such as tags, and titles, in the files themselves.
• I'll explore using f-spot and similar to see if they gain me anything, but I'll not commit to storing and managing my images with any single application.

I wanted to suggest a good package for image management -- but I forgot which one I last found, when looking for a Free alternative to picasa.

I do recommend trying picasa² -- it sets a nice benchmark for speed, and some features to expect. I suspect the Adobe-tools would be good inspiration as well.

However, the last package I came across looks very promising: Jbrout¹. It's written to keep tags in picture-data (no db), and makes use of lossless rotations etc.

Not sure if there's any support for digital negatives (DNG).

I also came across a Dr. Dobbs Journal article³ on indexing and searching images using lucene (the code is for .net). In general I'm leaning more towards a workflow/fileflow like the following:

1) Take pictures
-sort in camera (first round of deletions)
2) Copy to //.dng (theme usually
contains year, so others might prefer: year/theme or
theme/year.
3) Use a DNG-tool to preview fullsize. Second round of
deletions.
4) Tag/organize using something like picasa or jbrout/fspot.
5) Retreive/search using dedicated search/index
6) Edit/prepare for use -> save as new (ie:goto 4)

[1] http://jbrout.python-hosting.com/
² picasa.goole.com
³ http://www.ddj.com/cpp/210603684

Actually, I think the last program I settled on was digikam (http://www.digicam.org). I recently tested the version available in squeeze -- and while I only have a few small images on hand -- it seems really snappy -- and it handles digital negatives.