On Tuesday 16 Mar 2004 12:13 pm, David Matthews wrote:
On Tuesday, Mar 16, 2004, at 00:25 Europe/London, Paulo Jorge de
Oliveira Cantante de Matos wrote:
Is PolyML development dead? Is it safe to keep developing ML programs dependent on PolyML compiler?
Poly/ML is certainly not "development dead". I do continue to work on it and fix bugs but since this is a largely voluntary endeavour I need to fit this round my other work. My approach has always been to create a stable and efficient platform for writing Standard ML programs rather than to use it as a platform for experimentation. That tends to mean that new releases are infrequent and only when there is a significant change rather than producing a new release every few months. From feedback in the past I feel that this suits the users of Poly/ML better. ...
I endorse that completely. ProofPower compiles on both Poly/ML and SML/NJ but it's always been a lot more trouble keeping track of SML/NJ. The situation is particular bad on Mac OS X at the moment, where the only version of SML/NJ is much later than the version that's advertised as stable, and you need a source patch to work-around the differences between the basis library changes. As Poly/ML is leaner and faster for ProofPower, I much prefer it.
On Tuesday 16 Mar 2004 4:19 pm, Martin Ellis wrote:
On Tuesday 16 March 2004 12:13, you wrote:
On Tuesday, Mar 16, 2004, at 00:25 Europe/London, Paulo Jorge de
Is PolyML development dead? Is it safe to keep developing ML programs dependent on PolyML compiler?
Finally, Poly/ML is open-source. If someone wants a feature in Poly/ML badly enough they can always add it themselves.
...
If it seems sensible I'll add it to the source I maintain at polyml.org. The same goes for documentation. Although I maintain the version on polyml.org there's nothing to stop anyone else from distributing their version.
Apart from this, perhaps? "If you are downloading this in order to make it available to others, for example to install it on a server at your University or place of work, you agree that everyone who has use of Poly/ML will abide by this licence." [1]
I agree that this is a bit tricky. However, it seems to be part of the "agreement" rather than the licence itself. I suspect it would have no force whatsoever in law, since it's unenforceable and the "agreement" isn't a contract in law in any case (and nor is the licence).
I actually think this business of getting people to push an "agree" button is misleading. It's giving the mistaken impression that the licence imposes obligations on the user (as a contractual licence can do, e.g., to pay the licence fee). An open source licence is really just a public statement by a supplier of conditions under which a user can reasonably expect to use the software without risk of the contractor suing for misuse of IPR. I don't think it matters whether the user has in any sense "agreed" to the licence unless it's become the basis of a contract with a vendor.
It would be nicer to say something more practical. E.g. "you agree to make the terms of the licence known to everyone who has use of Poly/ML". This could easily be achieved by putting the text of the licence in the software somewhere. Of course, if the inclusion of this precise form of words was a CUTS requirement, David may find it hard to get it changed.
On Wednesday 17 Mar 2004 9:55 am, David Matthews wrote:
... The present licence was based on the BSD licence. It was drafted by a lawyer at Edinburgh and after a long delay it was finally signed by CUTS. A number of people looked at it particularly at LFCS in Edinburgh and made suggestions. The clause that seems to be causing problems was inserted largely to prevent the situation where a company developed, say, a new code-generator, and didn't make it available to everyone. On the other hand, GPL is too strong because it would appear to make it impossible for a company to develop software and distribute it with Poly/ML in such a way that they could retain their code as closed source. It's possible that the "Lesser GPL" that is used for libraries might be acceptable.
I think the GPL is actually much less restrictive than many people think (ProofPower is distributed under GPL, although I did agonise over this for a long time).
The relevant paragraph of the GPL is this:
2 b) You must cause any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof, to be licensed as a whole at no charge to all third parties under the terms of this License.
I don't believe that code compiled with a compiler counts as "derived from the compiler". The situation would, of course, be a bit less muddy if one had true separate compilation for ML (but even with gcc, compiled object code will contain compiled open source libraries, and I don't believe that the above paragraph means you can't sell binaries compiled with gcc).
Regards and apologies for this long off-topic diversion,
Rob.