On my long President's Day weekend I took shelter from the NYC cold at the recently-reopened Cooper Hewitt Museum on the Upper East Side.
The Cooper Hewitt is museum about design. On three floors and multple rooms the curators have arranged various sets of objects, often related to each other in some way other than the expanse of time. For instance, a Sumerian cuneiform tablet might be juxtaposed with an original iPhone. The focus is on the art of design and the ways that design helps us as humans. It is well worth a couple hour visit, though I believe even more additions will be made shortly.
My favorite piece was a display of pieces from the Free Universal Construction Set.
As a child, I owned some number of most of the major construction kits. Lego was by far the most dominant, but I also used Duplo, Playmobil, Lincoln Logs, Erector sets, K'nex, Hot Wheels, Tinkertoy, and Fishertechnik. Much of my playtime was spent combining these systems. Sometimes the scales could be frustrating but there was some satisfaction to send a Lego train through a western town made of Lincoln Logs while attacked by Playmobil giants.
The FUCS is a free and open set of 3D printable adaptors for a large number of toy systems, designed to make these toys interact with each other. Finally, a child (or child at heart) with a Makerbot can construct toys and systems that cross beyond the borders of corporate imaginations.
I got to thinking how these toy systems speak some internal language, but in order to work together the appropriate universal adaptors need to be supplied, and connected with other systems.
This is the same in data and computer science. To bridge two systems we need to agree on standard APIs, formats, and schemas. Oftentimes when one does not control the data itself in a section of the pipeline an engineer must define an appropriate adaptor to parse, transform, or display some set of information. Similarly, an interface needs to be implemented on the receiving end of a pipeline to do some sort of data load process. Good engineers live by defining appropriate, lightweight abstractions to complexity, allowing systems to communicate and lever each other's power, features, or datasets.
One example of a parallel in the software world are Protocol Buffers from Google. If you're not familiar with the concept, essentially JSON has become a standard data interchange format, except it's quite inefficient to read and write at high performance and since it doesn't compress well, is expensive to transfer. Additionally, the types are limited so we're left with a variety of add-ons and supersets to try to describe types like dates and times. Protocol buffers are not the solution to every problem, of course, but they do tackle the issue well on some fronts. Essentially you define a schema like file that describes your data, and then the system can generate an adaptor for your language of choice, creating a standardized way to serialize data to a binary format from a C program and read it in with something like Javascript. Protocol buffers are the abstraction that each subsystem can look to to communicate, allowing us to build polyglot systems.
The FUCS is a great, tangible representation of simple abstractions and interfaces for toy-engineering. Children today might get a better understanding of a system stack just by expanding their toy creations, and that is exciting.