On Thursday, August 27, 2015 at 1:10:33 PM UTC+2, Páll Haraldsson wrote: > > Strictly speaking, according to Stallman himself compilers do not affect > the license of binary produced by a compiler. However, the the C library, > that you usually incorporate, that is GPL, it would make the whole binary > GPL, if it were not for the GPL exception used. >
That's not the issue. There are some pieces of Julia, such as FFTW, SparseSuite, and one other I can't recall, that are under the GPL (and not the LGPL), so you can't even deploy them simply as a library. Without those, you're fine to deploy. > I wander what would be true for Julia code (that wouldn't at least use the > GPL parts of Julia). Yes, you are using Julia in the same process, but > would that mean all source code of yours would be automatically derived > software of Julia? That would make all Julia code GPL.. and doesn't seem to > make sense (any more than, Linux programs, derived software of the Linux > kernel, but it has an explicit, notice to that effect, mostly so people to > not assume that?) > > Maybe as Julia is unusual, has macros/metaprogramming, Julia should not be > considered as a generic compiler.. same argument should be made as for any > Lisp-compiler/system. > > On Thursday, August 27, 2015 at 10:50:50 AM UTC, Scott Jones wrote: >> >> We have some of the same issues, we will be using Julia code in a >> production environment, but at first, >> we will be hosting it, so we won't have the deployment issues. >> Deployment means that you will need to build your own version of Julia, >> without any GPLed code. >> There is a flag to do that. >> >> On Thursday, August 27, 2015 at 11:36:44 AM UTC+2, bernhard wrote: >>> >>> Dear all >>> >>> We have some Julia code which we would like to "deploy" client side. >>> Installing Julia (and possibly dependencies) on the clients machine(s) >>> should not be an issue. There are however two important restrictions we >>> face: >>> 1. The client data (on which our code will work) cannot leave their IT >>> environment. >>> 2. We want to protect our intellectual property (i.e. our code) at any >>> cost. >>> >>> Is this feasible today? Or is it partially feasible (i.e. in a way such >>> that decompliation is possible only to a certain extent). >>> >>> I must admit that I do currently not even know whether compilation is >>> possible (I think it is not in 0.3.10) and until now I always copied the >>> raw Julia code to other machines when I needed to calculated something on >>> other machines. >>> >>> Bernhard >>> >>
