Re: clickthrough license

[Per Bothner]

Mark Wielaard wrote:

> When I discussed this with Richard Stallman he said the main reasoning
> behind recommending a published book as primary reference when working
> on GNU Classpath is that if a book publisher can/has produced a book to
> explain the system,

He may have been thinking of books published by "authorities" (such
Sun employees or JCP experts).  Most Java books are written by random
people who don't have any special knowledge.  They are not published
to "explain the system" but to make money or burnish reputations.

> we can explain the system using source code

Example programs in a book are not sufficient to implement an API.

> (and publish this source code).

Huh?  Of course we can't publish the source code from a book!

> What I want to warn about is using anything that is not publicly
> accessible. If you need to actively consent to some click-through
> license then don't use it as your primary reference when working on code
> for GNU Classpath before contacting me (or the mailinglist).

I disagree.  It's a mistake to try to implement Java APIs without
looking at the *published* APIs.  It's difficult and unnecessary.

Something with a click-through license is publicly accessible, at
least in the sense of not being a trade secret.  I don't see how
clicking on Sun's documentation license tarnishes me in any way.

The agreement says that Sun grants me certain rights.  The only
requirement on me if I click "I agree" is to "indemnify, hold harmless,
and defend Sun" from my use of the specification.  However, there is
no prohibition on using the specification in ways not permitted by
the specification.  I.e. the license grants me permission to use the
specification in limited ways, and we can argue over what that
permission means.  However, I am not explicitly agreeing to not use
the Specification in other ways.  In such other ways, copyright would 
set the boundaries.

My point is that clicking "I agree" does not change anything I can
or cannot do with or for Classpath.  It does give permission to work
a compatible TCK-passing re-implementation (which is difficult to do
in Free project), but I have not agreed to not work on a non-TCK-passing 
re-implementation. (Sorry for the triple negative.)  And by reading
the specification I further Sun's stated goal of compatibility.

To expand on that last point:  By not using Sun's specification we
increase the risk of incompatibility, which *increases* the risk of
Sun suing under trademark law.  If we use Sun's specification, and
have a stated policy of treating any divergence as a bug to be fixed
(limited by resource availability) we're in a stronger position than
if we wilfully ignore the published specification.

> Also, I am happy to make sure specific cases get reviewed by FSF legal.
> But please make it a concrete case. "I am implementing <this> part for
> GNU Classpath and to make a real compatible implementation I need to
> access <this> information, which is only available if I accept the
> following <terms>. I looked here and there for alternative specs, but I
> am really stuck at this point."

No, we need a general review of the JDK specification license.  Some
people have been assuming that it conflicts with working on Classpath,
but I don't think there is an essential conflict.  A strict cleanroom
approach would keep implementors from directly accessing either Sun's
implementation or specification, but such strict separation is not
legally required, and we don't have the resources for it.
--
        --Per Bothner
address@hidden   http://per.bothner.com/