|
| 1 | +# TWG 2023-08-11 |
| 2 | + |
| 3 | +[Last meeting's notes](https://github.com/haskellfoundation/tech-proposals/blob/main/meetings/2023-07-14.md) |
| 4 | + |
| 5 | +# Meta |
| 6 | + |
| 7 | +Evie is resigning; we will attempt to recruit a new member at or following ICFP. Gershom will encourage applications there. |
| 8 | + |
| 9 | +# Project Updates |
| 10 | + |
| 11 | +# Proposals |
| 12 | + |
| 13 | +## Module Naming Conventions |
| 14 | + |
| 15 | +https://github.com/haskellfoundation/tech-proposals/pull/53 |
| 16 | + |
| 17 | +* Things in the threads that have not made it to the proposal text (e.g. `Experimental` as suffix rather than prefix to module names) |
| 18 | +* We prefer practical arguments over aesthetic ones |
| 19 | +* The mental model of the prefix vs suffix is that prefix suggests a logical relationship between everything there, while the suffix suggests that the contents are related to things that otherwise exist elsewhere. |
| 20 | +* There seems to be a consensus not to talk about the GHC API here |
| 21 | +* David will add a summary of this discussion as a comment |
| 22 | +* We think that sufficient stakeholders have been contacted |
| 23 | + |
| 24 | + |
| 25 | +# Other |
| 26 | + |
| 27 | +## Trees That Grow |
| 28 | + |
| 29 | +FYI Laurant left comment here https://gitlab.haskell.org/ghc/ghc/-/issues/21592#note_519447 |
| 30 | + |
| 31 | +Idea for potential proposal: getting the full benefit from Trees That Grow, resulting in the AST in a library outside of GHC that GHC users and a parser that's also outside GHC. The big blocker is the user of `FastString`, a GHC string type that features interning of strings. Potential solutions: replace FastString with Text or similar (performance allowing), parameterize AST over a string type. |
| 32 | + |
| 33 | +We suggest a really minimal proposal that says the goal is a separate package, and that it has one outstanding major concern to address. |
| 34 | + |
| 35 | +## Ways to Help Server Admin |
| 36 | + |
| 37 | +Can we pay a company to lighten the load for Haskell admins? |
| 38 | + |
| 39 | + * Hackage and to a slightly lesser extent downloads seem to be very sensitive due to risk of software supply chain attacks |
| 40 | + * Someone good at email config and migration would be really valuable |
| 41 | + * Most other things are nixified - this makes it much easier to maintain |
| 42 | + |
| 43 | +Two boxes we're concerned about that aren't Hackage or downloads: 90% nixified (www entirely, other than the Hoogle stuff). These aren't updated often. There are backups (done manually). |
| 44 | + |
| 45 | +Most value seems to come from finishing the migration to the new mailman. Most of that is done but there's some open questions. If we can find a contractor who's open to taking over two nix boxes and is skilled at email stuff who can do the mailman migration and then continue to take over the Postfix migration. |
| 46 | + |
| 47 | +Skills: |
| 48 | + * Mailmain |
| 49 | + * Postfix |
| 50 | + * Nix |
| 51 | + |
| 52 | +We could start with a contract for the mailman migration, then if they work out, move onwards to more. We should have the conversation. |
| 53 | + |
| 54 | +People to have in the conversation: |
| 55 | + * Gershom |
| 56 | + * Davean |
| 57 | + * Bryan |
| 58 | + * Ben G. |
| 59 | + |
| 60 | +Suggestion: Haskell companies will have more people who are into this. Haskell company that does Nix? John says that Obsidian did in-house email until recently, so they exist. |
| 61 | + |
| 62 | +Davean will put together a sketch of what's needed that we can use to get quotes. |
0 commit comments