Auto Metadata

One interesting part of the weekend photographic seminar I just got back from was the emphasis on metadata.

It's important to take great photos, but it's just as important to know how you got those results. In the days of film, A photographer would keep a shooting log, recording aperature, speed, lens, etc. A decent digital camera will record all that information for you as long as you know where to look for it.

Photography is one of the best examples of how useful metadata can be. The reason it is so useful is that it is unambiguous and immutable. Any metadata that has these qualities is easy to work with.

Pure Metadata

This is the simplest kind of metadata to work with and often the most useful. It usually has a direct connection to something tangible and concrete. Some types of pure metadata:

• File size
• File Attributes (resolution, color space)
• Physical Settings (ISO, aperature, shutter speed)
• Date*

Dates are very useful but only if they are relevant. If the date is not the actual shooting date it is actually less help than no date at all. This can happen with downloaded files or files that are saved to a new location.

Impure Metadata

File names are probably the least valuable indexing tool out there. They don't have to have any relevance to the content, an identical file can have different names and totally different files can have the same name.

Tags are good, but I don't find them as useful as they could be. Again, there isn't necessarily any correlation with the content, there is no standard set of tags or naming conventions enforced. But the primary reason it doesn't work for me is that it's a manual process. Any metadata that isn't automatically applied to all content is too much work.

Folder heirarchies are actually metadata and we use them every day. It's important to understand this when backing up or rearranging files because the heirarchy probably has meaning that would be lost if you moved those files.

The biggest problem with impure metadata, like the one's I've mentioned, is they can be very hard to normalize. There may be information attached as file names, tags and folder structures but it's all been manually assigned by what makes sense at the time. This manual metadata will always be very fragmented and incomplete.