[m-dev.] referenced papers

[Peter Wang]

On Mon, 24 Aug 2015 17:58:39 +1000 (AEST), "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> 
> 
> On Mon, 24 Aug 2015 16:12:33 +1000, Peter Wang <novalazy at gmail.com> wrote:
> > If anything, it should be a new git repository, e.g. "papers".
> 
> These references usually explain the design of a module. They play
> the same role as compiler/notes/*, the two differences being (1) they are
> often written by people outside the Mercury group, and (b) they are
> postscript or pdf files, not text. However, neither seems reason enough
> put people through more hoops to get at them.
> 
> While the mercury repo is of course usable for developers without
> the papers in it, we eventually decided to move tests into the same repo,
> after they were initially put into a different one. While the connection
> is not nearly as strong in this case, I would prefer not to repeat
> that process. I think the mercury repo should contain all the information
> sources that Mercury developers may need to understand the Mercury
> system. (Outside of basic background knowledge, of course.)

The connection is nowhere near as strong.  Tests and compiler notes need
to be updated with changes to the compiler.  Papers will never be updated.

There is the question of copyright.  Everything in the mercury
repository has been explicitly submitted for inclusion there;
some of the papers we reference will not.

The papers written by the Mercury group are already in the www
repository; I already have that checked out.  If there was another
repository, I would check that out as well or just download a specific
paper from the github page (rarely).  For me, there's no hoop.  I would
find it more annoying if the papers take significant space and end up
slowing down cloning of the repository.

Peter