From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-06-26 18:51:03
At 09:18 AM 6/26/2003, Howard Hinnant wrote:
>Since boost is a spring board for standardization of a library, I'm
>wondering if the boost license requires the copyright notice to follow
>for other implementations which follow the interface of the boost
>library, but independently develop the implementation?
A copyright, unlike a patent, just applies to the actual representation. So
unless another implementation actually made a literal copy of the Boost
code, the other implementation would not be a derived work of the Boost
code and so would not have to follow the Boost license.
This has already happened; the Dinkumware CoreX threads library follows the
Boost.Threads interface. But because they implemented from scratch, they
didn't have to reproduce the copyright or license.
Now you might ask, "what about the interface, doesn't the copyright cover
that too?" The answer is "no", as has been fought out in court several
times. Ask a lawyer for details, but interfaces themselves aren't covered
by copyright. The docs are covered, the header is covered, the
implementation, test cases, etc, are all covered, but not the conceptual
>In other words, if we standardize a boost library, will the library's
>copyright notice have to be in all implementations of that std::lib?
No, because the standard won't copy the actual library code. The standard
may copy actual prose wording from the proposal, and that in turn may have
been derived from the library's documentation, but the proposer will have
to assign the copyright to ISO on any portion of the wording that reaches
the actual standard document.
At one time some of us who wrote text for the standard actually had to sign
a copyright assignment to ISO, but then IIRC ISO just decided they
automatically became the copyright holder by some sort of international
treaty eminent domain, and stopped actually asking for a written
>Will the copyright need to appear in the standard itself?
Not normally. There are three exceptions in the current standard, if you
look at 1.10. So much of the standard wording came from three books by
Stroustrup, Kernigahan & Ritchie, and Plauger that a special deal was cut
with them. The three short paragraphs in 1.10 took a long time to
Another aspect of standardization is that anyone who proposes something for
standardization has to publicly announce if it is covered by a patent.
There was a recent court case (something about hardware memory IIRC) where
the proposer waited until after something got standardized, and then said
"you now owe my company royalties if you implement the standard". The judge
invalidated the patent. It would have been perfectly valid if the proposer
had pre-announced the existence of the patent.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk