I have a plain-text computer file with three sections: Sheets, "to be developed", and "in-camera".
A roll begins his life in the camera, and when I load the camera, I write which camera is loaded with what, in the "in camera" section. While I take the picture, I add short notes of the subject e.g.:
Minolta X-700: Astia 100F, Gianicolo (Garibaldi, Anita, faro), Lungotevere, Chiesa Nuova, Largo Argentina, Gesù, Grazioli (roll changed in front of Palazzo Grazioli).
When I unload the roll, it goes into the "TO BE DEVELOPED" section of the same file (copy and paste). The roll receives a double-letter code, e.g. FF. In the To be developed I now find this lines:
Roll FF: Astia 100F, Minolta X-700. Gianicolo (Garibaldi, Anita, faro), Lungotevere, Chiesa Nuova, Largo Argentina, Gesù, Grazioli (roll changed in front of Palazzo Grazioli).
When I develop it, the roll moves above in the Sheet section. It receives a definitive sheet number, and some definitive image numbers: e.g.
Foglio 23: Astia 100F, Minolta X-700. Gianicolo (Garibaldi, Anita, faro), Lungotevere, Chiesa Nuova, Largo Argentina, Gesù, Grazioli (roll changed in front of Palazzo Grazioli). Attributed numbers: 110487 - 110523.
The first two digits are the year of taking (2011 in the case of this roll) and the following numbers are a sequential attributed number.
The actual sheet is also marked with the sheet number and the attributed numbers. The stripes are loaded in the sheet in the order of taking. The sequence of sheets is coherent with the sequence of image numbers.
When I scan, I attribute to the raw scan the same number attributed to that image, e.g. 110487.dng.
When I develop the positive image, I attribute the same number to the image, e.g. 110487.tif.
By the same token, a JPEG derived from the same image would be 110487.jpg.
Scans from the same roll are all under the same directory. Directories are organized by year. E.g.
2011 (parent directory)
Foglio_23_Gianicolo_Chiesa_Nuova_Argentina_Gesù (directory)
Inside a directory, there are TIFFs and in a subdirectory the actual raw scans (DNG).
All images are described with short description, long description, keywords, and an approximate date of taking (I use Photomechanics, but then import the directories with Lightroom as well) using IPTC fields. Information is input once in DNG, and is then inherited by the other file formats.
Search on the entire stock are performed with Lightroom. Lightroom allows me to find the negative number, and from there finding the sheet, or the directory, is easy.
Digital images follow a similar logic, but they have no sheet and no plain-text file.
Image numbers begin with D so for year 2011 there can be an image which is called D110487 and has nothing to do with 110487. The two numbering sequences (film and digital) proceed in parallel.
Under the 2011 folder I would have for each day of digital captures a directory called, e.g.
2011_08_29_Paris_Luxembourg_Rive_Gauche_NotreDame
(directories of digital images are organized by day of taking, directory of film images are organized by film sheet).
That allows me to keep all images in the same tree directory structure by year, keeping side-by-side digital and film, without confusion, and to search by keyword any image, and see miniatures easily with Lightroom or PhotoMechanics.
Each image has its own IPTC fields. Lightroom creates sidecar files and/or populates a database, but I don't rely on that for archiving. I rely on images having their ITPC fields properly populated. A database can easily be created from actual IPTC fields of each image.