Project Harmony proposal adopted [LWN.net]

Harmony preview

Posted May 18, 2005 16:54 UTC (Wed) by mjw (subscriber, #16740) [Link]

Multiple people are trying to give a look into what Harmony will be like. Since it will most likely adopt existing technology around GNU Classpath instead of reinventing the wheel packaging Kaffe, GNU Classpath, Jikes and/or GCJ together will be a pretty good indicator. And it shows this will not be vaporware. See:

Many more examples on planet.classpath.org and on the harmony mailinglist. Happy hacking!

The C++ Standard Library

Posted May 18, 2005 17:26 UTC (Wed) by rriggs (guest, #11598) [Link] (8 responses)

Can someone with more insight into this care to explain why the GNU C++ standard library does not already fit the bill, or is not a better candidate for expansion in the desired direction?

My impression is that Rogue Wave's library was never any good. STLport and GNU's C++ standard library implementations have always seemed more mature and standards conforming.

The C++ Standard Library

Posted May 18, 2005 17:43 UTC (Wed) by ncm (guest, #165) [Link] (1 responses)

The GNU project might benefit from incorporating bits of this library, particularly bits that might free it from dependence on libc for locale file interpretation. Certainly the test suite will be a valuable addition.

The C++ Standard Library

Posted May 18, 2005 20:26 UTC (Wed) by nix (subscriber, #2304) [Link]

Without a copyright assignment, that strikes me as rather unlikely to happen. (Yes, even for the testsuite.)

The C++ Standard Library

Posted May 18, 2005 18:27 UTC (Wed) by JoeBuck (subscriber, #2330) [Link]

Rogue Wave's test suite would be a very useful contribution, and there might be other bits as well (assuming conflicting licenses aren't a problem). And I certainly don't want to discourage anyone from contributing code. But the project proposal reads like Rogue Wave, and its allies in Apache, want to replace GNU libstdc++ on the GNU/Linux platform, and take over its maintainance; after all, the two can't really coexist. I don't see any discussion in the proposal about how to deal with the LSB (though I have argued against the treatment of C++ in LSB 2.0). I also don't think that Rogue Wave will be able to afford to maintain this code for everyone indefinitely, and the people who are expert in this area and willing to make their work free software are busy working on libstdc++.

C++ is maturing. Apache shouldn't be in the business of replacing all the free software in the world with ApacheWare. Write clean, standards-conformant C++, and lean on the compiler providers to fix bugs.

The C++ Standard Library

Posted May 18, 2005 18:40 UTC (Wed) by ca9mbu (guest, #11098) [Link] (4 responses)

Yep. I too was surprised at this. The C++ standard is based on existing implementations - i.e. they only standardise what has already been proven out in the field. To this end, the Boost library (http://www.boost.org/) provides a whole slew of libraries for common, but not yet standardised features. Some of these libraries have already been proposed (and accepted) for the next round of standardisation. IMNSHO it would be more efficient to contribute to the existing Boost and libstdcxx projects, rather than go down the whole "Not Invented Here" road again.

From the proposal:

"Currently, C++ developers spend considerable effort porting code among
platforms, as compiler vendors are focused on backward compatibility rather
than cross-platform portability. There are other free implementations, but
none have the quality, license flexibility, or platform support necessary to
serve as a universal foundation for the C++ language"

I don't think GCC would agree about the lack of quality in libstdcxx, nor its portability. Such problems don't mandate a completely new library anyway; patches to the existing one would be adequate! So, shock horror, this would appear to come down to licensing issues...again!

The C++ Standard Library

Posted May 18, 2005 20:36 UTC (Wed) by nix (subscriber, #2304) [Link] (3 responses)

Licensing issues? There should be no problem using libstdc++ in proprietary code, thanks to the exception to the GPL that the appropriate parts of it are licensed under.

This feels to me like a mixture of Apache NIH syndrome and a possible desire to give Rogue Wave a new lease of life; as compiler vendors improve their standard library implementations, the space available for it has to be shrinking...

(Disclaimer: this has been a knowledge-free comment.)

licensing issues

Posted May 18, 2005 22:35 UTC (Wed) by bkoz (guest, #4027) [Link] (2 responses)

The idea of Rogue Wave donating their implementation to the open source community in a useful way, or otherwise joining with existing free software projects, is wonderful.

That's not what is going on here, however.

Their proposed license, the Apache License 2.0, is GPL incompatible.

http://www.gnu.org/licenses/license-list.html#GPLCompatib...

Thus, it's unlikely that this code will be able to be used by the GNU project. Which is really sad, because Martin Sebor is quite helpful and very talented: we'd love to have him and his great ideas contributing full-time to libstdc++. To be useful, Rogue Wave would have to do an assignment with the FSF and relicense the code. It seems obvious, at least to me, that Rogue Wave is not interested in doing this, or actually collaborating formally with existing free software projects. This is also sad.

For those who care, I suggest pointing this out to Apache, although without a doubt they are already aware of it. As Joe Buck pointed out here:

http://gcc.gnu.org/ml/libstdc++/2005-05/msg00137.html

this is likely something else going on here. Instead,this seems to be an attempt to have non-RW people work on a public version of their library, which RW controls and can then use in their commercial products.

Although perfectly within their rights, I find this end result much less than what it could have been. Sad.

licensing issues

Posted May 18, 2005 23:29 UTC (Wed) by nix (subscriber, #2304) [Link] (1 responses)

Their proposed license, the Apache License 2.0, is GPL incompatible. http://www.gnu.org/licenses/license-list.html#GPLCompatib... Thus, it's unlikely that this code will be able to be used by the GNU project.

Both the FSF and ASF admit that this is due to legal technicalities: now that the Harmony project's success is (in effect) riding on it due to the licensing of GNU Classpath, I expect some resolution. :)

However, the copyright assignment issue will probably still block it from significant use in GCC, I fear. (Joe would be able to speak more authoritatively here; I'm just an interested observer.)

licensing issues

Posted May 18, 2005 23:34 UTC (Wed) by nix (subscriber, #2304) [Link]

(Joe would be able to speak more authoritatively here; I'm just an interested observer.)

Sorry, brainfart, didn't notice who I was responding to. If anyone can speak authoritatively in this area, it's you ;)

(Apologies. I should not post when tired.)

One source for Harmony info?

Posted May 18, 2005 17:31 UTC (Wed) by b7j0c (guest, #27559) [Link] (1 responses)

Would be nice if there was a website/wiki/planet type source for all Harmony info...right now it seems to be just in the incubator mailing list.

One source for Harmony info?

Posted May 18, 2005 17:39 UTC (Wed) by mjw (subscriber, #16740) [Link]

Most of the hackers around Harmony (at least the people actually hacking on existing stuff) are aggregated on planet.classpath.org.

There is also a prilimenary wiki

Project Harmony proposal adopted

Posted May 18, 2005 22:09 UTC (Wed) by simon_kitching (guest, #4874) [Link]

Please note that Harmony has been approved as an *incubator* project.

That means it isn't an "official Apache project". It means that the Apache Incubator management committee has decided to allow a bunch of existing ASF members to create Harmony-related mailing lists, wikis and version control repositories on apache servers. And that there's *some possibility* of the results eventually becoming a real Apache project.

Even the expression "proposal adopted" is too strong in my opinion. Maybe "Project Harmony experiment allowed to start" is closer to the truth.

Note that I'm not being negative about Harmony; I'm just bothered that everyone seems to think that the ASF as a whole is somehow driving this project. The ASF isn't a company where things are *funded* by some management committee, and therefore every project that is started is "official". It's a community of programmers where different people have different interests. The ASF board tries not to let projects become "official" unless they have an active development community so Apache doesn't become a host for lots of dead code (*unlike* SourceForge!), and that's one reason for the existence of the "incubator" - to see if an experiment like Harmony will actually gather sufficient interest to live.