|
| 1 | +# TWG 2023-02-09 |
| 2 | + |
| 3 | +[Last meeting's notes](https://github.com/haskellfoundation/tech-proposals/blob/main/meetings/2023-01-12.md) |
| 4 | + |
| 5 | +## Present |
| 6 | + * Laurent |
| 7 | + * Davean |
| 8 | + * David |
| 9 | + * John |
| 10 | + * Luke |
| 11 | + * Gershom |
| 12 | + * Hécate |
| 13 | + |
| 14 | +## Agenda |
| 15 | + |
| 16 | +### Project Updates |
| 17 | + |
| 18 | +#### Advisory DB |
| 19 | + |
| 20 | +#### errors.haskell.org |
| 21 | + |
| 22 | +Davean: how many visitors? |
| 23 | + |
| 24 | +David: Unknown; we're hosted on GitHub pages and don't have something like Google Analytics on the page. |
| 25 | + |
| 26 | +Gershom: What about hosting it on the haskell.org hosting? We can turn off IP addresses in logs if worried about GDPR. |
| 27 | + |
| 28 | +Hécate: What about users in Iran and Cuba? |
| 29 | + |
| 30 | +Gershom: Someone was blocking our CDN. |
| 31 | + |
| 32 | +David: No objections to moving to haskell.org hosting. It would be nice to have stats about which pages are most popular and which 404s are most popular. |
| 33 | + |
| 34 | +Davean: We can add awstats that extracts stats from a log in a cron job before the logs are deletes. |
| 35 | + |
| 36 | +David: Who does this? |
| 37 | + |
| 38 | +Davean: Send SSH key for uploader and for David to get SFTP access. |
| 39 | + |
| 40 | +Gershom: Here's the haskell.org scripts: https://github.com/haskell-infra/www.haskell.org/tree/master/.github/workflows |
| 41 | + |
| 42 | +Timeline: |
| 43 | + 1. Davean does awstats, David sets up parallel uploading |
| 44 | + 2. Then we debug the new host, when it works, transition DNS. |
| 45 | + |
| 46 | + |
| 47 | + |
| 48 | +#### Timing |
| 49 | +Can we move it 30 minutes later? |
| 50 | + |
| 51 | +Works for all present. |
| 52 | + |
| 53 | +### Open Proposals |
| 54 | + |
| 55 | +#### Standard Library Reform |
| 56 | +https://github.com/haskellfoundation/tech-proposals/pull/47 |
| 57 | + |
| 58 | +The proposal was just updated immediately prior to the meeting. There seems to be an emerging consensus that this is a good idea. |
| 59 | + |
| 60 | +Remaining work is small details, read-through, then un-draft and submit for a vote. |
| 61 | + |
| 62 | +David asked for specific requests for help, rather than general ones. Perhaps a synchronous read-through on a call could be possible? |
| 63 | + |
| 64 | +Laurent volunteered for a synchronous read-through with John. They can also discuss Haddock. |
| 65 | + |
| 66 | +#### IDE/Test Integration |
| 67 | +https://github.com/haskellfoundation/tech-proposals/pull/46 |
| 68 | + |
| 69 | +David was supposed to send a quick email to get a Davean/santiweight conversation going, but failed to do so. Will do. |
| 70 | + |
| 71 | + |
| 72 | +#### Maximally decoupling Haddock and GHC |
| 73 | +https://github.com/haskellfoundation/tech-proposals/pull/44 |
| 74 | + |
| 75 | +We should probably close this right now, not because it's a bad idea, but because it exceeds current time resources. |
| 76 | + |
| 77 | +David: Is it worth spending some of the foundation resources to make it happen? |
| 78 | + |
| 79 | +Gershom: could this understanding be converted to a GSoC proposal? Even if it's not done in GSoC, having it in that format is a great way to get volunteers in other contexts. |
| 80 | + |
| 81 | +Laurent: Can definitely do this. Not sure whether it's too much work. Will try to make the proposal by Feb 12th. |
| 82 | + |
| 83 | +Gershom: Will give feedback. |
| 84 | + |
| 85 | +#### Other Business |
| 86 | + |
| 87 | +##### `ghc-prim` cleanup |
| 88 | + |
| 89 | +Laurent: On John's proposal, Bodigrim suggested looking at people to move from `ghc-prim` to `GHC.Exts` from `base`. Can we all work together to follow reverse dependencies from `ghc-prim` to point them at `base` instead? |
| 90 | + |
| 91 | +Davean: We can't do this for everything, some programs just need `ghc-prim`. |
| 92 | + |
| 93 | +Gershom: Using Hackage reverse dependencies? |
| 94 | + |
| 95 | +Laurent: yes! 14k packages transitively depend on `ghc-prim`. |
| 96 | + |
| 97 | +We discussed a variety of packages that use it, but where we don't think that it should. |
| 98 | + |
| 99 | +Davean: most uses of `ghc-prim` are in fact incorrect, becuase using the unboxed type is equivalent to the boxed one, and programmers are copy-pasting. |
| 100 | + |
| 101 | +John: we need to establish clear guidance as to what the best practice is here, rather than relying on our fractured oral histories. |
| 102 | + |
| 103 | +Gershom: There's a give and take. The guidance will come from changing things and seeing what happens and can't be entirely _a priori_. |
| 104 | + |
| 105 | +Can we scale this up? Instead of Laurent doing the work directly, what about doing the work but documenting it underway and recruiting further recruits as we go? The `prim` cleanup crew proposal? |
| 106 | + |
| 107 | +Laurent: This is the idea for the proposal. Two parts: |
| 108 | +1. Identify low-hanging fruits (`ghc-prim`, what else?) |
| 109 | +2. Write the guidance |
| 110 | + |
| 111 | +##### Hackage new account approval |
| 112 | +Gershom: The Hackage approvals team gets burned out. Can we automate the process more? We need to avoid spam (manual measure). The current one is arduous - we need to increase the size of the team. We don't have an exact answer. |
| 113 | + |
| 114 | +David: What about an arxiv style vouching process? |
| 115 | + |
| 116 | +Gershom: Proposals welcome - can we write that up into a worked-through proposal? GSoC or otherwise? |
| 117 | + |
| 118 | +Davean: We need to keep the current one for now. What about publishing the requirements on the page instead of sending them as an email so people can do the work up-front? |
0 commit comments