There’s a lot to be learned from taking the assorted files we all have laying around on our computers and taking them apart, seeing what makes them tick, and then breaking them. First and foremost, doing this is a chance to recognize that, although we may look at an image file or text file as simply an image or text, they only appear as such when viewed through a lens provided by our computers. Those files, however, are not those things. At their most basic, they are set of electromagnetic states of certain physical spaces on your hard drive (or electrical states of your solid state drive, if your computer uses one of those). In an intermediate sense, the computer uses more basic lenses to view those states as code, and that code may need several layers of interpretation by your computer before you see it in the form it was intended to be viewed in. This should remind us that, although we often imagine a dichotomy between the physical world and the digital, all digital objects are merely physical objects which are being viewed in a certain manner. We can achieve similar understandings of the physicality of digital items by placing a magnet near one’s hard drive, provided one does not mind loosing the further use of that hard drive. We should also be reminded that the appearance of these digital objects is not static, but depends on what lenses they are being viewed through, and we are also reminded that, although two image, text, video, or sound files of different types may appear similar to us, they are actually quite different.
While it may sound technically complex to some, glitching a file can actually be quite an easy thing; many of us even manage to do it by accident. To demonstrate, I will use 3 different types of image file (with the hopes of demonstrating how these are different), as well as an audio file. Those trying this themselves should make sure to make copies of all the files they intend to glitch and preferably cordon them off in a separate folder, to make sure they don’t accidentally break a file they didn’t mean to. For the images, I’ve selected a .jpg (a preliminary design for an unbuilt class of USN warships), a .png (a shot from the opening of a recent, award winning television show), and a .gif (a motivation penguin). The specific images have been chosen in the hopes that they will prove memorable and to show that one can learn from glitching whatever images one has on hand.
To begin with, the .jpg in an unaltered state:
Next, we will convert this file to a text file. To do this, you will need to rename the file, so that the extension (the 3 letters at the end of the file name after a period, which exist to tell the computer what type of file this is and what program it should use to open it) says .txt instead of .jpg. If you cannot see any file extensions, open file explorer options from the control panel and ensure all options pertaining to extensions have been set correctly. When the file has been converted to a .txt file (a common type of text file), it should then be possible to open it with a word processor. Some word processors, such a Microsoft Word, may recognize that the file is not what it appears to be, but wordpad and notepad will both suffice. I will be using Notepad++, a version of notepad with extended functionality. The result looks like this:
Note that although this is mostly gibberish, the “Photoshop 3.0” appears in the first line, and “8BIM” appears several times in the first few lines. Removing these will not actually change the image. On the other hand, deleting several full lines, particularly from the very beginning or end, can easily result in a file which cannot be opened. More judicious removals of one or two lines here and there from the beginning of the body of thefile resulted in this:
Copy and pasting segments of text at the center of the document, meanwhile, added this to the image:
By doing this, one can see the methodical way in which the data which the image represents is stored; if code from the top of the file is removed segments of the image are also removed, starting from the top. .png files, on the other hand, are less straightforward. Here is an unaltered .png and the same file as text:
Once again, the Photoshop name appears, but beneath it, we see a collection of metadata, intended to provide us with information about what this image is. Most of it is incomprehensible to those who are not more technically skilled than me, but the “history” section is interesting, as it contains a date, presumably when this file was created; this date does coincide with when the show was airing.
If we start deleting segments from the body of this image, we see a much different result:
The new image contains not only the missing and correct segments we saw in the .jpg, but also an area which is highly distorted. This indicates that the file does not store data in a purely linear fashion, pixel by pixel, but stores some information about one area of the image in one part of the file and other information elsewhere. Also worth noting is that the blog did not allow me to display this image, so the print screen function was used to replicate it.
.gif files are interesting because, although they are an image format, they allow for simple animations to be displayed. Here is one accompanied by the same file as a .txt. Note that you will need to click the .gif for it to play.
Once again, there is a small amount of comprehensible text, this time in the form of “GIF” at the very beginning of the file. Editing the file results in a much shorter animation playing, with a small amount of distortion of the image:
As you can see, this also results in the .gif playing automatically, for reasons which are beyond my technical skills to understand. We can see from this that even a small amount of glitching can result in large changes to a .gif file.
As I am unclear of the legality of simply uploading an unaltered music .mp3 file to wordpress, I have elected to use music which all readers should be able to access through Youtube, specifically, the song Sonntagsfahrer (Sunday Drivers) by the Ostrock band Puhdys. If readers are interested in making their own glitches of the same song, an .mp3 can be purchased from Amazon, and otherwise, it can be listened to here: https://www.youtube.com/watch?v=ACZU4wTd5uE
As a file, the song can also have its extension changed from .mp3 to .txt, which looks like this:
Once again, the file starts with a small amount of metadata, such as that this file came from Amazon. Editing this file as a text document can produce surprising results; it often takes the removal of large swaths of text to produce a change in the song, usually in the form of segments simply disappearing, rather than being distorted. I will refrain from posting my own examples, again in the interest of not attracting the attention of amazon’s lawyers, but one can easily see this by editing their own .mp3s.
I hope this has proven instructive, as to the nature of the files we often interact with. If your own experiments with glitching files produce any interesting results, please share them below.
These metadata are a reminder that all things digital are representational in nature. This is easy to forget this because our minds are so very good at abstracting and understanding the meaning behind the representations. A computer user might look at a photograph or utilize a document, but these items are merely representational of the real thing. In fact, no real thing can exist on a computer.
Think of Michel Foucault’s This Is Not A Pipe, chapter one, Two Pipes. A drawing of a pipe is captioned “this is not a pipe.” Of course, this seems silly at first, because it is, in fact, a pipe. Yet, as Foucault explains, it is not a pipe but merely a representation of a pipe. That is the key to the digitization and the key to understanding these metadata.
Just as the song or a picture or a document on a computer is only a representation of any of those items, so is the metadata a translation of that representation. What is key here is that the information is intact, only inaccessible to users. There are certain formats for different data sets and once they have been translated out of those formats, they become unusable until translated back to their initial representative state.
This begs the further question of whether or not the storage of an item is complete only when it is useful to people?