"David A. Wheeler via rb-general" <[email protected]> writes:
> I personally prefer a "meaningful" datetime stamp, since that provides > additional information. What may be missing is consensus on what a "meaningful" timestamp actually is, for different common scenarios. Is that something we can build consensus on and make recommendations about? Alas, this is a complex area often with no real right or wrong, and the right trade-off may depend on context. That can explain the various (arbitrary) SOURCE_DATE_EPOCH patching going around. For example: * Date in man pages - there is no real guidance on this. Reasonable choices: 1) Time of last modification to file. Does not work well for auto- generated man pages based on multiple sources which themselves may be auto-generated. 2) Time of last public release of upstream. 2b) Time of generating upstream tarball, including snapshots. 3) Time of first creation of the manpage. 4) Time of last public release of packaged version of the tool. 5) Don't include a timestamp at all. There are probably more choices. I suspect each of these are in use. Doing some archeology to what dates were used in old releases of Solaris and BSD's would be an interesting data point. * Date used in Texinfo manuals. Autoconf has logic to set this based on mtime of *.texi but this interacts poorly with playing of mtime of tarball content. * Date in po/*.pot files. Gettext v1.0 improves this, but it is very new. * Other common examples? /Simon
signature.asc
Description: PGP signature
