Hello again! I believe that I've found a substantial issue in the Java port of libidn (that's not in the C code). Maybe splitting off the Java part into its own repository wouldn't be the worst idea anyways.
Simon, if you're still not opposed to this, could you please a sibling repository (https://codeberg.org/libidn/libidn-java) and hand me the keys to that particular castle? I'm https://codeberg.org/guus on there. I've got a split from the original project lined up locally (while retaining history). I'll then address the issues in the new repository, rather than muddy the waters further in the multi-language project. Kind regards, Guus On Wed, Jul 15, 2026 at 12:48 PM Simon Josefsson <[email protected]> wrote: > Guus der Kinderen <[email protected]> writes: > > > Hi Simon, > > > > Although I do believe that a separate project for the java code has > > considerable benefits, I wonder if it's worth doing that, given the > > maturity of the entire project. It's decades old, and the foundation on > > which it is based is frozen. Does splitting it up now make sense? > > Probably not - but I wanted to mention the possibility. After removing > the *.JAR file from the 'make dist' tarball, most of the Java stuff > doesn't cause a lot of problems any more. It does add complexity to > normal releases, though, since I need to have Java etc installed and > updating javadoc pages takes some time. > > > Then again, from a cursory glance at libidn2, that doesn't have _any_ > Java > > port right now? Establishing a distinct Java project is perhaps something > > that could be applied to both. > > Libidn (including the Java port) is IDNA2003 so really ancient. I would > be surprised if there weren't multiple IDNA2008 Java implementations > around? If not, doing one would be nice, but it is not a trivial job. > I wouldn't want to add that to Libidn2 though, let's keep it separate if > you or anyone wants to work on that. > > /Simon > > > So many pet projects though... > > > > What do you think? > > > > - Guus > > > > On Sun, Jul 12, 2026 at 8:55 PM Simon Josefsson <[email protected]> > wrote: > > > >> Guus der Kinderen <[email protected]> writes: > >> > >> > Hi Simon, > >> > > >> >> No worries - getting things published definitely has priority. > >> > > >> > I pressed 'publish' moments ago. \o/ > >> > >> Great! > >> > >> > Though your reply did make me look more closely at something, and I > don't > >> > entirely like what I found. > >> > > >> > What struck me is that between 1.15 and 1.44 there are a great many > >> changes > >> > to the project, and essentially none of them touch the Java port. > C-code > >> > surely is quite different from the Java code, and the code is punned > to > >> > Unicode 3.2 so has little reason to move - but still: that's an awful > lot > >> > of non-Java-change releases. > >> > >> I don't think there has been any significant algorithm-related changes > >> since pretty much forever though. > >> > >> > It makes me doubt my earlier assumption that the Java port isn't a > >> > frozen-but-faithful twin of the C library; I'm now more inclined to > >> believe > >> > that it's an independent ~2004 implementation by a different author, > that > >> > may have drifted. > >> > >> That is right -- it was a contribution, and what little Java I ever know > >> it has probably fallen out of fashion. > >> > >> > So perhaps, some of those releases probably *should* have touched the > >> > Java side? > >> > >> Hard to tell -- but not certainly. > >> > >> > I've still published 1.44 today. What's on Central is 1.15 from 2009, > >> > unsigned and without sources, so this is a clear improvement > regardless. > >> > But it maybe should not be the last word. What do you think? > >> > >> Improvements are always welcome, so maybe your update will spur some > >> interest. Possibly all of the Java code could be moved into a separate > >> git project, I think ~20 years ago multi-language projects made sense > >> but today I think it is just a pain to have to coordinate independent > >> implementations in a single release. > >> > >> /Simon > >> > >> > Regards, > >> > > >> > Guus > >> > > >> > On Sun, Jul 12, 2026 at 5:28 PM Simon Josefsson <[email protected]> > >> wrote: > >> > > >> >> Guus der Kinderen <[email protected]> writes: > >> >> > >> >> > Hi Simon, > >> >> > > >> >> >> Yes I am happy if you can sign artifacts and make the upload. > >> >> > > >> >> > Thanks! > >> >> > > >> >> >> Are they reproducible? > >> >> > > >> >> > Not as things stand. The jar and javadoc tooling both embed > >> timestamps. > >> >> > Fixable, but I'd rather leave it out of scope for now and get 1.44 > >> out. > >> >> > > >> >> > What I built: the Java code from the 1.44 tarball, unmodified, > >> recompiled > >> >> > with Java 8 for compatibility. > >> >> > >> >> No worries - getting things published definitely has priority. I > don't > >> >> know about Java reproducible artifacts, or if anyone in the > Central/Java > >> >> community cares. > >> >> > >> >> >> I suspect Java people aren't using *.JAR's from GNU tarballs, but > get > >> >> >> things through Maven or other sites, though. > >> >> > > >> >> > That's exactly right, and the current situation isn't very good: > >> Central > >> >> > has only 1.15 (2009) and 0.6.5 (2006): > >> >> > https://mvnrepository.com/artifact/org.gnu.inet/libidn People can > >> route > >> >> > around it with private repositories (e.g. > >> >> > > >> > https://igniterealtime.org/archiva/repository/maven/org/gnu/inet/libidn/ > >> >> ), > >> >> > which is unsigned and unofficial and not a real fix - and likely > >> outdated > >> >> > too. Not having it in Central will effectively make it unavailable > to > >> >> most > >> >> > developers. Getting 1.44 onto Central should sort it. > >> >> > >> >> Yay! > >> >> > >> >> /Simon > >> >> > >> >
