Create 2023-01-12.md

Author: david-christiansen

meetings/2023-01-12.md

@@ -0,0 +1,77 @@
+# TWG 2023-01-12
+
+[Prior meeting's notes](https://github.com/haskellfoundation/tech-proposals/blob/main/meetings/2022-11-10.md)
+
+## Present
+
+* Hécate
+* John Ericson
+* David Thrane Christiansen
+* Luke McCartney
+* Jack K
+
+## Agenda
+
+### Project Updates
+
+* Advisory DB
+    * No progress right now
+
+* errors.haskell.org
+    * Now supported by GHCup and Stack
+
+### Open Proposals
+
+#### Haddock/GHC Decoupling
+
+* Meeting tomorrow to discuss it between Hécate and Laurent
+
+#### IDE/Test Integration
+ * TODO: facilitate contact between Davean and proposal author, because Davean has a useful and relevant prototype of aspects of the proposal
+
+#### Standard Library Reform Proposal
+https://github.com/haskellfoundation/tech-proposals/pull/47
+
+* Where did the time estimate come from?
+    * Based on work experience, done conservatively
+* John: How do we feel about the first step being a big boring, and having that be the only funding proposal?
+    * David: If the first step is all that happens, have we contributed value? Is the world better, worse, or neither?
+        * John: Mostly neither
+        * Gershom: Even if end user experience is not better, is there a better experience for core library devs and GHC devs?
+        * John: Hopes so
+        * Jack: Did the overall structure change?
+        * John: The big update is adding this "get the PR over the finish line". Also, if Safe Haskell is removed, this gets cheaper and easier.
+    * David: Do we have feedback from GHC devs and CLC about whether their lives are better with just this first step?
+        * John: Recall hearing back from Bodigrim, who was fairly neutral, and Ben, who was positive
+    * Gershom: If we did Step 1, and also a little bit of Step 2 (which can be incremental) then we would get an immediate benefit because we'd have a few things that could be upgraded without reinstalling GHC.
+        * John: Technically yes (reinstallable and upgradable) but there are some gotchas for running TH natively and ABI compat. It's unclear how much leaks into the ABI of GHC itself. But if a stage1 compiler is fine, then yes.
+        * Gershom: What's the ABI compatibility issue?
+        * John: TH and parts of GHCi work by loading the user program into the GHC process, and they can do whatever. If a library that GHC uses were upgraded, then that could lead to problems.
+        * Gershom: this is surprising. If I use Bytestring 11 on a a GHC that shipped with 10 and used TH, will that cause a bug?
+        * John: It's only if you're using these things in the Template Haskell
+        * Gershom: You need to use the bytestring that GHC is linked against? That's a problem today, right? You're your own worst salesman.
+        * John: Yeah, that's right
+        * Gershom: If we do step 1 and part of 2, then many things are actually improved.
+        * The best demonstration would be to take out things that we regret having put in `base`
+    * Gershom: Where does that leave the proposal?
+        * John: There's a big "future work" section at the end that isn't formally asking for money
+        * Remaining steps: 
+        * Gershom suggestions:
+            * Move prior work and motivation later
+            * Put the actual proposal higher up in the document - today one must read to 6.1 to see the suggestion, and not until 7 does it say that there's no more! the first para should contain the actual concrete proposal, and then elaborate on it later (appendices, prior art, related work)
+            * Fill out alternative prelude section
+        * John: still needs to address David's comments
+
+### Misc
+
+#### Tooling thread
+
+* John has made progress towards a reinstallable RTS, and triaging the details of the various configuration settings and options
+* This should make the bindist metadata problem less daunting and more structured
+* This should be generally useful for tooling
+
+### Meta
+
+ * How did the emailed reminders work?
+ * Decision: try again another month
+