From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 25184 invoked by alias); 23 Apr 2002 12:46:32 -0000 Mailing-List: contact java-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: java-owner@gcc.gnu.org Received: (qmail 25150 invoked from network); 23 Apr 2002 12:46:23 -0000 Received: from unknown (HELO alpha) (194.236.130.162) by sources.redhat.com with SMTP; 23 Apr 2002 12:46:23 -0000 Received: from oskar ([192.168.1.2] ident=mail) by alpha with esmtp (Exim 3.35 #1 (Debian)) id 16zzfd-0005qf-00; Tue, 23 Apr 2002 14:45:21 +0200 Received: from usel by oskar with local (Exim 3.35 #1 (Debian)) id 16zzd9-0006MQ-00; Tue, 23 Apr 2002 14:42:47 +0200 Date: Tue, 23 Apr 2002 07:25:00 -0000 To: Bryce McKinlay Cc: James Williams , java@gcc.gnu.org, gcc@gcc.gnu.org Subject: Re: Attention GCJ devel team Message-ID: <20020423124246.GA24351@oskar> References: <001e01c1e797$c91c2f30$c47831d2@computer> <3CC38926.90804@waitaki.otago.ac.nz> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3CC38926.90804@waitaki.otago.ac.nz> User-Agent: Mutt/1.3.28i From: Oskar Liljeblad X-SW-Source: 2002-04/txt/msg00321.txt.bz2 On Monday, April 22, 2002 at 15:10, Bryce McKinlay wrote: > >I am currently working on a tool that will generate class stubs complete > >with javadoc from javadoc specifications. In your FAQ a person > >mentioned that "Considering that new Java APIs come out every week, it's > >going to be impossible to track everything." I believe the tool I am > >developing may reduce this development challenge for you substantially. > > > >From my perspective the value of this would be that when a new > >specification came out, the api converter could be run providing a clean > >framework complete with all the new and deprecated api's and then the > >intergrator could copy the existing code from the current libgcj > >implementation into the new framework. > > This sounds like an interesting and useful tool. However, I don't know > whether or not we can, from a legal perspective, generate code > directly/automatically from Javadocs. I suspect we'd have to ask the FSF > legal people whether or not this is safe. I wrote a program similar to what you describe some time ago. It is called DeDoc and can be downloaded from http://freshmeat.net/redir/dedoc/1802/url_tgz/dedoc-0.12.tar.gz. It reads HTML (generated by javadoc from java source files) and writes source files for each class encountered. I don't know if anyone actually uses/used it... The same question came up that time - is it legal to use code automaticly generated from Javadoc [in GPL programs]? I wrote RMS to ask about it, and he forwarded my email to Eben Moglen. I quote from the letters I received: Eben Moglen writes: The act of removing the entity definitions from the documentation and reusing them for reimplementation is the same whether an automated process is used or the copying is done by hand. So if the single question is whether running the DeDoc routine is legally more problematic than doing the work by hand, the answer is no. From the copyright point of view it makes no difference whether the data was compiled by a very automated or less automated process. The larger question is whether the copying that is taking place, no matter how, could conceivably be held infringing. Numerous factors enter into this, including what proportion of the HTML documents are being copied, whether the entity signatures could satisfactorily be characterized as functional rather than expressive, what proportion the signatures bear to the size of the finished work, and even what the identifier names are. Some measures of ordinary prudence should be taken. Where identifier names do not require to be externalized for compatibility with existing library linkages, they should be changed. Where other aspects of the signature are not similarly required, they too should be reconsidered. Before release we should review the code a little more closely to make sure that what could be reinvented rather than copied has been. Here's more from Eben Moglen: OL> The signatures information is the base for about 95% of the output of OL> DeDoc (counting bytes of code.) This may seem much, but know that OL> what DeDoc generates is not usable immediately. It still takes OL> heavy (re)engineering efforts to complete the library which DeDoc OL> provides the code-base for. Once the library is completed, the signatures OL> will probably be less than 2% of the finished work. Then we are not going to worry about the matter for now. The proportion of signature to other HTML input to DeDoc is also probably low, because of the text comprising the rest of the documentation. So our case against an infringement by copying claim is essentially a fair use defense: small amounts of copying from the original document, functional in character, representing a small proportion of our finished work. Under US copyright law this would be a sound position and I am comfortable that we could defend it. Let's be sure to check at the later stages of the project, to be sure that our assumptions turned out to be right. Oskar Liljeblad (oskar@osk.mine.nu)