[rb-general] Heading toward functional compilation by adding hidden inputs?

[Ximin Luo]

John Gilmore:
> The reproducible-builds team is trying to evolve compilation to become
> more like a mathematical function -- having defined inputs and
> outputs, and always returning the same output from the same inputs.
> Removing hidden inputs like the date, the local directory name, the
> host system type, etc.
> 
> But what I find bizarre is that the way you are going about it is by
> inventing NEW hidden inputs -- rather than evolving toward removing
> more and more of the hidden inputs from the set of build tools.
> 
> Perhaps it comes from frustration with the feature-oriented mindset of
> the tool developers -- if you ask them to REMOVE a feature, they are
> dubious, while if you ask them to ADD a feature, they are more willing
> to listen?
> 
> I would try to resolve that tension by defining reproducibility as the
> feature you are trying to add.  (Not by defining "ability to secretly
> control timestamps" or "option to override debug paths" as the
> feature.)  As a former tool developer myself, I would be happier to
> add reproducibility than to add any of those quirky obscure features.
> Maybe there are others like me, if you can reach them.
> 

talk is cheap, show me the code

-- 
GPG: ed25519/56034877E1F87C35
GPG: rsa4096/1318EFAC5FBBDBCE
https://github.com/infinity0/pubkeys.git