I have no real understanding of this, but following the bread crumbs, I found this in Pijul's site (my emphasis below):
> In that paper, and in Pijul, the definition of a file is slightly expanded, to included these cases, and “artificially add” all pushouts of all possible patches to the space of files.
> The resulting structure is a graph of byte chunks (which can be for instance binary blocks, lines, or smaller parts, such as indentation).
> The relation between them, described by the graph edges, is the order in which they appear in the files. This graph acts as a “cache” of applied patches.
> In that paper, and in Pijul, the definition of a file is slightly expanded, to included these cases, and “artificially add” all pushouts of all possible patches to the space of files.
> The resulting structure is a graph of byte chunks (which can be for instance binary blocks, lines, or smaller parts, such as indentation).
> The relation between them, described by the graph edges, is the order in which they appear in the files. This graph acts as a “cache” of applied patches.
From: https://pijul.com/model/#pushouts-and-distributed-algorithms
The paper is they mention is the same one in GP's linked article.
A Categorical Theory of Patches - https://arxiv.org/abs/1311.3903