The poetics of network latency

Over the past decade, I’ve had the privilege of being involved in several projects that explore the limits of the World Wide Web as a support for live multi-sited performances as well as the potential for network latencies to become expressive features of such real-time performances. Instead of being understood as the enemy to be overcome in order to achieve a “live-like effect,” network latencies become a necessary, even desirable feature to be interpreted as a feature of each performance.

The question of the expressive range of network latencies –whether on the server or client end matters little– has been on my mind with respect to Ghost and Mallarmovie: two recent experiments with the Zeega platform. Each is a tightly constructed sequence of slide animations deliberately designed to push the limits of the medium of the database documentary, layering gifs and stills with such a high degree of density that the play time of individual slides as well as the animations that compose them becomes only loosely programmable. 

Visual rhythms, playtimes and sequencings thus come into being as a function of a delicate balance: the interplay between user inputs, the bandwidth available on the client side, the Zeega server, and the servers feeding up the source files that are being stitched together in real time. 

So how different is each performance of a single piece? Just how variable are the pace, sequence of actuation, and couplings between components? I knew from experience that, under differing network conditions, the variations could be considerable, especially if the end user either slowed down or forced the pace. So, I thought it might be more interesting to document another sort of limit case: the same piece, in this case MALLARMOVIE, played five times at roughly the same pace on the same device from the same location on five different occasions during a single day.

So here goes:

First take:

Second take:

Third take:

Fourth take:

Fifth take:

I haven’t timed anything with precision, but here are some elements that may be worth tracking comparatively:

–the opening spiral sometimes loads with the railway track and rolling dice gifs, but does so for only several cycles before stopping; in other cases it remains static or rotates uninterruptedly

–the pace of paging gifs (slides 3, 5 and 10) varies slightly from one play to another; in some cases the gif actuates belatedly, rather than simultaneously, with the loading of the slide; the speeds are highly variable, particularly in the case of slide 5 and 10

–the play or rotation speeds of various gifs can be seen to vary from one take to the next; this appears to be the case with the loading icons on slide four, of such flickering words as “jamais” and “au fond d’un naufrage” on slide seven, and the weather fronts in slide 15

–the blinking red light in slide 10 almost never repeats itself and is a sensitive indicator of the real-time stitching process with highly variable advances and lags