In short, it's a simple grammar used for describing more complicated grammars. But let me explain further.
I mentioned in my previous post that I'm trying to create my own programming language. Like human languages, programming languages have their own syntax and grammar. This is the thing that goes wrong when you get a syntax error.
There's a whole lot of theory behind how a grammar can be described. One of the most common ways of specifying the grammar for a programming language is Backus-Naur form. This is basically a metagrammar (or metasyntax): a grammar for describing grammars. And there are lots of variations thereof.
Most programming languages use a formal description like this to define their syntax. But they don't let you change it. The grammar of the language is usually hardcoded into the parser itself.
One of my language design goals is to let the grammar itself be a data object available to the user. This way, you could extend the basic syntax that I provide in whatever way suits the program you are trying to create. In fact, I want to take advantage of that capability myself. I want to only hardcode a fairly simple grammar into the language, then load the more complicated grammar from a data file.
What will be the format of that file containing the full grammar of the language? Well, it has to be readable by the simple grammar that is hardcoded into the language. (Of course, I could break it up into more steps than that, but for now I think one step will be sufficient.) So this grammar file must present the full grammar of the language using a simpler metagrammar.
This, then, is the bootstrap grammar: a simple metagrammar used to define the more complex grammar of the actual programming language.
I do have a draft of the bootstrap grammar right now. In fact, I have attempted to create a file that describes the bootstrap grammar using the bootstrap grammar. (As if this whole discussion wasn't meta enough already.) This is actually useful for testing the parser: I can check that the grammar produced by the file matches the grammar used to read it.
Of course, as this is a hobby project that I only do in my spare time, progress is slow. I have not yet completed a successful "round trip" of the grammar like the kind I described. But I am encouraged by the results so far, and I think there is a good chance this will eventually work.