[Python-Dev] Building on OS X 10.4 fails

[Thomas Wouters]

On Wed, Jan 18, 2006 at 02:25:03PM +0900, Stephen J. Turnbull wrote:

>     Anthony> It sounds like configure needs to grow a test to detect
>     Anthony> that a "libreadline" it finds is actually the crackful
>     Anthony> "libedit" and refuse to use it if so.

> FYI: Real libreadline is GPL, and rms made a point of forcing
> (Aladdin-licensed) Ghostscript to remove stanzas from the Makefile
> that allowed linking to it as a user option.  Ie, this particular pain
> in the neck is deliberate FSF policy, to encourage use of the GPL.

[...]

> As long as the link to fake libreadline succeeds and the resulting
> program works identically to one linked to real libreadline, he has no
> complaint.

I don't think this applies to Python. The Aladdin license isn't
GPL-compatible, but the current PSF license is (according to rms himself.)
(Only, IIRC, 1.5.2-and-earlier, 2.0.1 and 2.1.1-and-later, not 1.6, 1.6.1[*],
2.0 or 2.1.) The Ghostscript check-for-readline is a case of "you are still
linking with readline, even when you aren't actually linking" -- but this
isn't a problem for (most versions of) Python, because rms said it isnt. As
long as the resulting Python binary is only covered by the GPL-compatible
PSF license, the GPL and no GPL-incompatible licenses, any form of linking
is fine, even configure-time-not-building linking.

[*] 1.6.1 is the release that contained the license change rms and his
laywers wanted, but that itself doesn't make the license GPL-compatible. It
apparently allows derived works from being GPL-compatible, though. 2.0 and
2.1 derive from 1.6, but 2.0.1 and 2.1.1 derive from 1.6.1.

I'm sure it makes sense to someone.

Go-figure'ly y'rs,
-- 
Thomas Wouters <thomas at xs4all.net>

Hi! I'm a .signature virus! copy me into your .signature file to help me spread!