Where should the type information be: in tags and descriptors 418

Aw heck, I'll byte too :-)

The first time I saw this, circa perhaps 1970, it was some CAD data interchange files. All the tables of "things" had current memory address as the identifier for each "thing". For "thing" think line, arc, spline, circle, etc. both definition and instance of same. The files came from two different CAD systems. "Your job, Hank, is to create a converter that will port data both ways between those CAD systems."

The first thing I realized was that those "memory addresses" were NOT memory addresses, but were just uniqe names for the objects. Thinking of them as memory addresses does not lead to any useful outcome :-)

Sometime much later I was given an almost identical task, except that the files had been created by C++ programs. The programmers involved really really wanted to talk about "pbuttivating" and "activating" the data, and really really wanted to think of the memory addresses as pointers. So reading the data file involved "pointer swizzling". I never got even one of them to understand that the memory addresses were NOT file pointers, even once they were "swizzled". They were just arbitrary unique names, and thinking about them that way made life simple.

The example I used was Betty, our next door neighbor. Sure, I could point toward her current location and mention that there was an object at that location with property Gender: Female and property Height: 66 inches and property HairColor: Light Brown. But it is so much easier to use her name. If you want to talk about where she is located you might say "At Home" (an indirect reference to another table database) or might give her address (the name of the location of her house) or might provide a geographic locator (township, range, section ...another name of the location of her house)

So once a pointer has been placed into a file, it is no longer a pointer, but instead becomes a name. "Hank's rule" says that you should never ever put binary data into a file, but should put a text representation of that binary data into the file instead. All interchange files will be text. If you run across one that is not, just pretend the "pointers" are names.

--

... Hank