Saturday, February 27, 2016


I just spent a little time pulling together an album of photos from our recent vacation, and fully pulled the plug on Picasa, using native Google Photos interfaces throughout. This involved photos from a handful of sources, taken at various times.

I found that, in general, the preferred method for organizing things was to let Google do it. While you can organize things into albums if you like, Google does not make it frictionless or even particularly easy. This fits in with a paradigm I've been observing for some years now, the first instance I noticed being Apple's "Library" model for iTunes, iPhoto, and so on. I found it particularly curious with iTunes, because music actually comes pre-organized, and the very first thing iTunes does when you shove an album of music into it is to throw the album structure away. It is now just a bunch of tracks which you might or might not be able to re-assemble into the original structure if you search carefully.

The paradigm that vendors want you to use is to simply shovel all your stuff into an enormous heap. Then the vendor will let you search it, and will organize it to a degree for you. Photos get sorted by whatever date the software pulls out, and optionally by faces or whatever.

This is wrong.

The function of a computer is not to store all my shit in an enormous heap. The function of a computer, the cloud, whatever, is to impose order on my data in a way that makes sense to me. No, your ideas for how that order should be imposed are probably wrong. They might be a useful approximation, but they're wrong.

Automatic organization needs to be an assistant, not the end-game. You need to have a clear notion of a "working set" of stuff I am monkeying around with. If I upload a photo now, I need to be able to find it, trivially, and organize it relative to other objects. Sorting it to Jan 1, 1970, because the date field is all 0s is not an acceptable model. I need to be able to select things and put them into boxes, temporary boxes and permanent boxes. I need to be able to put boxes inside other boxes.

Directory structures for filesystems, heirarchical structures of data, are near-perfect models of how to get it right. And we're chucking them away as fast as possible in favor of this new thing.

The reason here is that order matters. I don't want to simply share all my pictures with the world, or even with my friends. I don't want a book made out of some automatically selected thing. Order, organization, reflects my idea of how my stuff ought to look, how it ought to be presented, how it should be shared, how it should be viewed. If you make a few books, regardless of the kind of content, you will quickly learn that the content is a startlingly small part of the problem. It is the structure, the organization, that is difficult and important.

Now, I'm a weirdo. I don't simply want to create a single linear stream of every picture I take and share it with the world, and I am fairly confident that this is exactly what the Average Consumer wants to do. Which is unfortunate, and represents some sort of further dumbing down of everything.

Quite apart from the Average Consumer is a technical problem. Big searchable heaps are wonderfully scalable. If you want to store a trillion things and have multiple copies across multiple servers and disk drives and whatnot, maintaing any organizational structure other than, basically, tags attached to the things, is difficult. Google, not being in to actually solving difficult problems and being extremely in to search, loves the enormous searchable heap. Apple has just lost its way, and everyone else just follows those jokers, and here we are.

The dominant model, the big searchable pile, destroys order and prevents the creation of order.

There was a quotation running around a decade or so ago, I think attributed to Tim Berners-Less, "flat text is just never what you want" which is exactly wrong. Flat text is almost always what you want. If you can't order your thoughts into a relatively coherent linear flow with some footnotes and cross-references, then your thoughts are stupid and incoherent. Order, structure, matters, and it enrages me that we're moving at such a dizzying pace into a world in which our thoughts, our ideas, our songs, our pictures, every product of human imagination, is simply heaved into a big unstructured, more or less sort of searchable heap. Tag something wrong, and it's fucking gone. Oops. Sorry.

Monument, please take note.


  1. Bravo! I don't always agree with you, but this time I do - in spades.

  2. Hi Andrew:

    That's why I like Lightroom. I have a hierarchial file system as the base, then add keywords to enhance searches. I have a base directory called "Photos", then under that subdirectories for each year. In those, I have directories for each event, like a vacation, and under that, any subdivisions I feel I need. Example:
    Let's me search on that, or through keywords (if I've been there twice for example).

    I don't like the "big pile of stuff" either

  3. its all about layers of abstraction and the interface that you, or an user, wants to see.
    your example of the directory structure for file systems is actually very relevant. the 'directory structure' that you see is what is presented to the user, it has nothing to do with how the bytes (or even the files) are physically stored on the storage media.
    The advantage of 'databaseing' your data, in the very broadest sense of making the individual items searchable in some form, is that you can then present the data/files/pictures in any way you like independent of the storage form.
    Obviously the big issue is auto capturing the relevant metadata so that the images/files/whatever can be found again.

    Having said that i totally agree with you that dumb interfaces (iTunes etc) are far from optimal and often don't match the 'flavour' of the content they are trying to present.