[rb-general] Paper rough sketch about reproducible builds...
jcappos at nyu.edu
Mon May 15 22:21:53 CEST 2017
On Mon, May 15, 2017 at 11:40 AM, Ximin Luo <infinity0 at debian.org> wrote:
> Hi Justin, could you add me to the repo? My github username is infinity0.
> As a note for others, on Debian you need to install the "texlive-science"
> package for "make" to work.
> I'm not sure about this part:
> > [..] We propose a more precise goal, specifi-
> > cally dictating *where* reproducible build bugs should be fixed
> > rather than just focusing on the goal of making builds re-
> > producible. [..]
> This is nice as an ideal goal, but I didn't think we're ready for it yet.
> I don't see this expanded in the rest of the paper though. Did you have
> some more concrete ideas towards it?
I don't really have concrete ideas, but the paper really needs to be about
*something*. I thought that it was interesting to claim it is bad to
either change the build environment (i.e., running in a VM / container) or
doing some post processing steps. We could argue for changing the source
code of the build systems, which seems the much harder way to solve the
> Regarding SOURCE_DATE_EPOCH, it should be noted that it's somewhat of a
> "compromise" solution. The general motivation is described here:
> You should probably also mention buildinfo files, and how they differ from
> things like Dockerfiles which I briefly touched upon in the other thread,
> "Regarding Zero Install manifests". The general motivation is described
> and in slightly-more precise theoretical terms here:
These sound like great things to add / correct. Do you want to take a stab
at the text? I could do so also, but it will be a few days...
(If you do edit, feel free to just push your changes to master when you are
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the rb-general