Ian Jackson <ijack...@chiark.greenend.org.uk> writes: > I am against the version number change. Version numbers, and particular > version number bumps, in protocols like this are a very heavy hammer. > Where present, they should be bumped only when necessary. They are > necessary only when a tool which is written to the old format will *do > the wrong thing* with the extended format.
> If a reasonable tool written to the old format will either do a > plausible thing, or reject the input, if fed the new format, then it > is a bad idea to bump the version number. > That is the case here. My preference is that at the end of this change we have two fields, a License field that includes the full license text and a License-Identifier field that has just short identifiers (including now using License-Identifier in places where the current format uses License without an accompanying license paragraph). That change will require changing all parsers to understand the new format. That change is somewhat painful, but avoiding it is going to be more painful in the long run. It was a mistake in the original specification to allow one field to have two different syntaxes in a very context-sensitive way, and now we would be adding even more complexity to the decision between those two syntaxes. I believe we should just fix the original specification bug and be done, painful though that may be, and allow parsers to not carry this complexity load forever. > TBH I think even having a version number at all in the > machine-readable copyright format is quite possibly a mistake. That's other entirely valid approach: to just set the format in stone, never admit a backward-incompatible change, and live with the initial decisions, warts and all, that cannot be fixed without breaking backwards compatibility. That's what MIME ended up doing. But I don't think our install base is so large as to require that, and I do think it's feasible to fix all the parsers. The reason MIME went down that path is because it was not reasonable to fix all the parsers. -- Russ Allbery (r...@debian.org) <http://www.eyrie.org/~eagle/>
