[libvoikko] HFST licensing

Krister Lindén krister.linden at helsinki.fi
Wed Jun 2 14:51:51 EEST 2010

In the current HFST libvoikko implementation we have used the full HFST 
library, but this is only intended for development and testing purposes. 
In the final version we will have a tailored addition to the HFST 
optimized-lookup code for the speller and hyphenator. We hope to have 
this addition available by the end of the summer.

The HFST library and command-line tools are mainly intended for creating 
and manipulating lexical transducers. The tools create transducers which 
only have licensing restrictions that are dependent on the original 
lexicon material given to the tools.

Our HFST optimized-lookup software avoids possible licensing problems as 
the optimized-lookup format and runtime code is fully within our control 
and has explicitly been given multiple licensing options. However, the 
Apache license is the one that we most often refer to as being the most 
liberal if someone also needs the code for commercial purposes.

The HFST library which communicates between the various implementation 
libraries is also in our control and can be explicitly released under 
multiple licenses. However, a user will normally only use the HFST 
library with one particular implementation library, so if someone wishes 
to include the HFST library as such in an application, he will also have 
to check the license of the chosen underlying implementation library.

To be explicit, the HFST library contains or will contain the following 
implementations in the near future using all of them simultaneously 
while conveniently converting transducers between the implementations in 
order to take advantage of different implementations that are efficient 
for different tasks:
- OpenFST (weighted transducers, Apache)
- SFST (unweighted transducers, GPL)
- HFSTOL (weighted or unweighted transducers, multiple [our own code])
- foma (unweighted transducers, GPL)
- fsm2 (weighted transducers, GPL)
- ...

In addition, the HFST optimized-lookup (HFSTOL) library can be used as a 
stand-alone library. This library only has a limited set of functions 
dedicated to fast lookup in precompiled transducers, e.g. morphological 
analysis, base-form reduction, spelling, hyphenation, ...


Harri Pitkänen wrote:
> On Wednesday 02 June 2010, Krister Lindén wrote:
>> The HFST library can very well continue to be released under GPL2 as a
>> whole.
> Does this include implementations/openfst-1.1 too? It is under the Apache 
> license. Common interpretation of the licenses is that you can't release 
> Apache licensed code under GPL2 unless the authors of the Apache licensed code 
> (in this case Google and at least two independent contributors) have 
> explicitly agreed to it.
> There is some disagreement whether this is a correct interpretation but at 
> least FSF and Fedora seems to have written licensing guidelines that consider 
> these licenses incompatible.
> Harri
> _______________________________________________
> Libvoikko mailing list
> Libvoikko at lists.puimula.org
> http://lists.puimula.org/listinfo/libvoikko

More information about the Libvoikko mailing list