For small changes, there can be a fourth way between the fix and the fork: maintaining a _patch set_ for the official version. One ecosystem where this happens often is the Linux kernel.
The practicality of this depends on the nature of the changes and the rate of upstream conflicts. But if you can automate the application of the patch, maintaining a "fork" can be a much lighter affair. It would be nice if patching was handled as a first class dependency mechanism by high-level build systems (e.g. npm, cargo, maven, etc.)
The practicality of this depends on the nature of the changes and the rate of upstream conflicts. But if you can automate the application of the patch, maintaining a "fork" can be a much lighter affair. It would be nice if patching was handled as a first class dependency mechanism by high-level build systems (e.g. npm, cargo, maven, etc.)