Sending to the mailing list instead of to Herbert alone...
On Thu, Sep 7, 2017 at 2:20 PM, Herbert Valerio Riedel
On 2017-09-07 at 18:16:39 +0200, Mario Blazevic wrote:
Btw, here's an old commit which updates the class diagram to this effect for the report:
https://github.com/hvr/haskell-report/commit/ 339ea257ee8b0451fbba388480566efac6ecbbd3
Ha, I wasn't aware of that repository.
I set up the hvr/haskell-report fork[1] shortly after I migrated and set up the haskell/haskell-report repo back in 2015 to serve as an "updated" inofficial Haskell201x report...
While looking through the report it became apparent to me that more updates may be needed, and that a new Haskell Prime committee was needed because such an inofficial Haskell report wouldn't provide the desired authority of a properly produced language standard, and you know the rest... :-)
That looks farsighted for sure.
We agreed today to move the report itself to the https://github.com/haskell/rfcs/ repository.
Ok, so how does this change the procedure described at
https://github.com/haskell/rfcs/blob/master/README.rst#succe ssful-proposals
?
I think the only necessary change is to the strangely worded clause - No one is appointed responsible for actually implementing the change, in particular neither the shepherd nor the author of the proposal. I'd go with some alternative wording like - The successful proposal should include a complete delta to the text of The Haskell Report that can be automatically merged.
And what is the intended relationship between the haskell/rfcs and the haskell/haskell-report repos?
Should we move the build system around it as well? I'd say probably not, leave the haskell/haskell-report repository the canonical one and update it from haskell/rfcs/ once we're ready to publish.
Well, depends... the build-system is a bit incomplete as it only tests that TeX still builds, the intention was to provide a CI system which publishes its draft aftifacts somewhere for convenient previewing. And if I understand this correctly, you intend to have RFCs be accompanied by deltas to the report in the same repository; and if that's the case I think the build-system makes a lot of sense to duplicate in the haskell/rfcs repo.
I'm not familiar with the build system, so I'll trust your judgement on this. The only reason for my earlier choice is that haskell/haskell-report sounds like a proper cannonical place for the official Haskell Report, much more so than haskell/rfcs.
If the report was written in reStructuredText we could simply use something like the readthedocs.org service. But since it's LaTeX, we have to do a little bit more work to publishes ("deploys" in newspeak) .pdf drafts somewhere else, but it's doable.
I can take care to set it up, if it's clear what kind of CI/CD we want.
Is the current publishing system really that difficult? To my grizzled ears, this sounds like you're fishing for a volunteer to translate LaTeX to ReST. I'd actually be willing to do that, as I have plenty of experience with text transformations, but I'd need a buy-in from everybody.
I wish GitHub made it possible to symlink files in two repositories like this.
I wouldn't worry too much about that... we can cross that bridge when we're close to a report worth publishing :-)
Cheers, HVR