Re: Google Summer of Code 2015

[John MacFarlane <jgm-TVLZxgkOlNX2fBVCVOL8/A@public.gmane.org>]

+++ Ivan Lazar Miljenovic [Feb 22 15 14:12 ]:
>On 22 February 2015 at 10:27, John MacFarlane <jgm-TVLZxgkOlNX2fBVCVOL8/A@public.gmane.org> wrote:
>> +++ John MacFarlane [Feb 21 15 15:15 ]:
>>>
>>> +++ Matthew Pickering [Feb 20 15 21:45 ]:
>>>>
>>>> Dear list,
>>>>
>>>> I have created a ticket on the Haskell.org GSoC idea list for one of the
>>>> ideas which John suggested[1] would be suitable. It is important that
>>>> that
>>>> projects are fairly freestanding (not requiring intimate knowledge of the
>>>> architecture) and are selected based on utility to the Haskell community
>>>> as
>>>> a whole.
>>>>
>>>> If anyone has any comments, feel free to edit the proposal or reply here
>>>> and I will make the changes.
>>>>
>>>> https://ghc.haskell.org/trac/summer-of-code/ticket/1660#ticket
>>>
>>>
>>> It's a good proposal.  Some further motivation:  pandoc's current
>>> Markdown parser is not very efficient.  It even goes exponential on some
>>> inputs, which is not good for web use.
>>>
>>> I've already developed algorithms for parsing CommonMark efficiently,
>>> without backtracking.  They are so much more efficient than what pandoc
>>> currently does that even the JavaScript implementation of commonmark is
>>> 3-4 times faster than pandoc, and the C implementation is 30-40 times
>>> faster.
>>>
>>> So I'd hope for a 10X speedup with a rewrite.
>>
>>
>> Further thoughts on this.
>>
>> The best GSOC projects benefit the whole community (infrastructure).
>>
>> So, it might make more sense to write a standalone CommonMark parser
>> library with a liberal (BSD) license, that could be used in other
>> Haskell projects (e.g., potentially, in some future version of Haddock).
>>
>> If this is made extensible, pandoc could simply use this library.
>> But the library would also be available for other purposes.
>>
>> Pandoc itself won't ever be relicensed.  But I think it would be a good
>> idea, if contributors will agree, to dual license pandoc-types BSD3/GPL.
>> (There are only a few commits by people other than me, so this is
>> feasible.)
>
>If pandoc-types is licensed BSD3, is there any real need for it to
>also be licensed GPL (especially since Cabal doesn't really understand
>dual-licensing)?

Probably not.