[[[ To any NSA and FBI agents reading my email: please consider ]]] [[[ whether defending the US Constitution against all enemies, ]]] [[[ foreign or domestic, requires you to follow Snowden's example. ]]]
In general terms, I think the VC data will do the job. What worries me is what exceptions there are. For instance, there are git commands for asking which functions were affected by a certain commit, and finding commits that affect a certain function. I suppose these work well in the usual cases. Are there unusual cases they don't handle right? For instance, do they work for macro definitions in C? What about structure or class definitions? File-scope variable declarations? There might be problems in cases that don't occur to me. Perhaps we should ask all GNU developers to try out how effective the VC facilities are. This means that for a certain period of time, everyone should continue maintaining change logs, but try to avoid referring to them. When you need information about past changes, you'll try to get that information solely from the version control system first. If and when the version control information is less helpful than the change log, you should report that. If this test shows that there is little loss in having no ChangeLog file, we would stop maintaining ChangeLog files. If it turns out that the lack of a ChangeLog file is a pain in some cases, we would investigate why and how. Then we would try to improve the version control facilities, or our way of using them, to eliminate this disadvantage. I think a proper test would have to last a year. Some cases arise rarely but may be quite important when they occur. -- Dr Richard Stallman President, Free Software Foundation (https://gnu.org, https://fsf.org) Internet Hall-of-Famer (https://internethalloffame.org) Skype: No way! See https://stallman.org/skype.html.
