Subject: Re: For Approval: BSD License, PostgreSQL Variant
From: "Chris Travers" <>
Date: Wed, 10 Oct 2007 18:01:56 -0700

> I don't see how the approval of a license would produce such an outcome.  I
> am more than game to follow what our process states, which is to discuss and
> recommend to the board the approval, rejection, or further discussion of a
> license.  But the submission of *this* license is unlikely for the OSI to
> drop its /position/ (not a mere pretense) that the definition of open source
> itself is based on peer review.

Approval of all permissive licenses currently in use would reduce
confusion over what is/is not open source.  For example, is ICU (i18n
Components for Unicode) open source?   The license is substantially
similar to the MIT license in structure but the "New BSD" license in
terms, but that specific wording has not been approved.

If that is the case, then we need to take the view that *all*
compliant license should be approved, and that we look specifically at
structuring the list in a way which better controls the proliferation
of licenses.  More thoughts on that below.

> I would like this list, and the board, to consider the proposition that we
> could define a trivial compatibility licensing cohort, namely, any licenses
> that permit either relicensing under substantially similar terms (such as
> SimPL does for the GPL) or permits arbitrary copying, modification, and
> redistribution along with any other licenses (subject to a limitation of
> warranty for the initial contributor(s)) and is subject to no further
> restrictions be treated as a member of the cohort license group.

I think that this is a good idea, but I would suggest that be taken it further.

Basically, most of these licenses consist of minor variations in
wording and restrictions compared to other licenses.  This includes
the ICU license which is the wording of the MIT license, slightly
changed, with an additional non-endorsement clause.

I would suggestion instead having approved variants.  This has the
following benefits over the "approval as a category" approach:

1)  Cuts down on actual license proliferation.  A genuinely new
license has a higher barrier to entry compared to a variant.  Also a
variant does not get its own listing.  It only gets mentioned on the
page of the other license (in this case, this would be similar to the
ISC license, and a variant of that).

2)  Better organized, so easier to find information one is looking
for.  If I look up the MIT license, and I get to see the (which
strikes the word "sublicense") and ICU licenses (which strike the word
sublicense, and adds a non-endorsement clause) as well, then that is a
good thing.

3)  Streamlined but not eliminated discussion.  Discussion would
simply be over changes to the licenses that were being approved, not
the entire license itself.  Therefore, if the Aferro GPL were ever
submitted, the network access provisions alone would be the topic of
discussion and it would not get its own license page (in fact, we
could consolidate the GPL and LGPL into one listing too).

4)  An ability to remove some retired and non-retired liceses from
listing pages (such as the Intel Open Source License) due to the fact
that they are just variations on other license.

However, if this structure were not adopted, I would accept your proposal.

> Moreover, licensors who are willing to specifically agree that their license
> is, in fact, if member of a permissive cohort take the step of permitting
> relicensing under other licenses (perhaps with the limitation that reference
> to the initial license be maintained so that credit goes where credit is
> due), the cohort would collapse to a single quantum licensing state, making
> the proliferation question moot.

I can't be certain what you mean by a "single quantum licensing state."

However, if you are proposing that there is essentially one listing
which includes all approved licenses of this category, then that would

> Licenses like the GPL (versions 2 and 3) are cohort sinks: they permit
> others to relicense under their terms, but not to be relicensed under other
> terms (except by decision of the author(s)).  Our license proliferation work
> attempted to identify logical (if not legal) base equivalencies, and to the
> extent we have, it's up to those who choose substantially similar terms to
> agree that via some mechanism of preserving licensing history (a form of
> attribution!) they can agree to consolidate around the popular ones (or a
> neutrally-named cohort class).

In general, I agree with the idea.  The questions are merely ones about details.

Best Wishes,
Chris Travers