Pathing in the D ecosystem is generally broken (at least on windows)

Artur Skawina via Digitalmars-d digitalmars-d at puremagic.com
Sat Sep 26 16:47:42 PDT 2015


On 09/26/15 23:58, Laeeth Isharc via Digitalmars-d wrote:

>> Given the DMD licensing situation, __nobody__ will (or should) even look inside the DMD repo for info. Especially that

Note that the above is not what I actually wrote, but has been altered
with no mention of this fact.
It's hard enough to convey tone via email; such manipulations don't help.


> He's entitled to his view, but normally one is taken more seriously if one makes a reasoned argument for a strong view (which he declined to do in that previous thread).  Prudence is a virtue, but it's not quite the same thing as blanket aversion to all possible risks - each must judge for himself, but advising others like this goes quite far.

It's not advice, but a statement of fact. Well, the `(or should)` part
/is/, but it was parenthesized for a reason - it's not the main point,
but only a preemptive response to any potential "but they should" reply.

Obviously, "nobody" in this context does not literally mean "nobody",
but nobody from the set of people with an interest in the subject that
might potentially create open source or otherwise differently licensed
works. The latter subset can in theory be the same as the whole set
(it will be smaller in practice, yes). Considering that this discussion
is about an apparently undocumented file format that Manu would like to
see supported in a differently licensed work (LLVM) and thinks that
Walter and/or DMD is a good, or even unique, source for info about, then
yes -- _nobody_ (that would like to use the information to indirectly
incorporate in into LLVM) will look for it inside some other proprietary
compiler. At least, they are _not_supposed_to_, and really shouldn't.
Even without malicious intent it's too easy for the result to be
similar enough that somebody can claim it's a derivative work. And even
when such a claim is obviously bogus, you do not want to have to deal
with it.
Hence, as it appears that the code in question is boost licensed,
(re-)publishing it in a way that would limit the "contamination"
concerns might help Manu's cause, and does not require Walter do much
more than a git clone+add+commit+push. Convincing a LLVM developer
to support a file format that's documented in a single boost licensed
file is going to be much easier than suggesting that they obtain the
info from a non-free non-redistributable compiler source from another
vendor. And by "much easier" I mean "possible", because the other option
simply isn't (and shouldn't).

Now, I don't know if the info in that file really is as unique as Manu
says, plus because of this thread it already became much more accessible,
so it's possible that the issue has been already solved. But every
other `free-but-entangled-with-non-free` part of DMD has the same problem.

"Let's look inside works we can't legally use, just in case there's some
usable part inside" is not a viable option. Really.

artur


More information about the Digitalmars-d mailing list