Below is the main part of the PM I sent to Jennifer Donelson following her question ("what would CMAA taking over Gregobase entail exactly")
As you can see, I don't object to CMAA taking over (not that I have a say in the matter), but I would rather first attempt to manage Gregobase as an informal community.
From a technological standpoint, Gregobase's software is exceptionally well-made for a 2010s amateur developer, not too badly made for a 2010s professional, and extremely outdated by today's standards, which makes the addition of new features quite difficult and risky. I audited its security and maintainability and it is bad, but not catastrophic. On this front, first, there needs to be a technical debt resorption effort / technical modernization and refactoring, then, there are a few bugs that have been reported and need fixing, and finally, long-standing features that have been asked for by the community in the comments of Olivier's blog pages. These include an API, a better versioning system, discussion pages for entries, better CRM/ticketing, and many more. Gregobase's code is hosted on GitHub with Olivier as its sole GitHub admin, which in my opinion is one of the factor that discourage community contribution to the code. The development effort could be crowd-sourced (Jakub Pavlik and myself have volunteered to add features that interest us as frequent users of Gregobase). There are also evolutions required depending on choices regarding data, see below.
From an infrastructure standpoint, Gregobase runs a single instance on Olivier's personal server (
selapa.net). This is okay for software on which no one relies, but for something like Gregobase it is very bad, because it means that:
- when Olivier changes the code, it gets put into production very quickly, sometimes introducing regressions. Fortunately, Olivier almost never changes the code (hence the technical debt).
- if Olivier forgets to renew his server subscription, Gregobase disappears.
To fix those two issues, there typically needs to be a testing instance which can be broken and fixed constantly without causing issues with the main instance; and a frequent backup of the main instance.
Gregobase is also currently vulnerable to DDOS attacks where thousands of bots spam the website with requests and the website stops working. This is typically avoided by using Cloudflare as a middle man between the user and the website: Cloudflare checks if you're a human the first time you connect to Gregobase from a new computer, and then remembers that you're not a villainous bot. However, Cloudflare has its drawbacks: it sometimes (rarely, but not never) breaks and with it breaks all the websites it protects.
From a data standpoint, Gregobase now consists of about one half of entries transcribed from modern-era chant books (published 1870 and later), which were its initial target, and one half of entries that are not linked to a source, being either restitutions from manuscripts, adaptations of existing pieces into non-Latin languages, or personal compositions. The "incipit" and "usage" (i.e. office-part) fields are well-populated which makes it relatively easy to find a piece if you know what you look for (e.g. communion antiphon Quotiescumque, Vatican edition). However, the rest of the metadata (Version, Source, Commentary, CantusID) is very inconsistently populated. There is also an estimated 10% duplication rate (2,000 duplicates out of 20,000 entries). I volunteered to work at fixing it, hence why Olivier gave me admin access to the database in order to make my work easier. There is currently an explosion of entries consisting of adaptations into English, Spanish, and a variety of Vietnamian dialects. The website was not really made with this in mind and, while this influx of contributions has little impact on the workings of the website, it does get heavier and heavier to navigate and search.
From a governance standpoint, there is no governance. Anyone (bots included) can setup an account and add or edit scores with no oversight at all, except me who attempts to clean up the worst parts of the mess once in a blue moon. There is a short explanation of what each field means (Incipit, Version, Commentary, etc.) but no community standards and no mechanism to enforce them. Several decisions need to be made about the handling of personal compositions and the handling of scores in foreign languages in order to both enable Gregobase to provide clean score data to its "consumers" but also a welcoming hosting space for its "providers".
So, in summary, Gregobase needs: two separate servers (or one beefy one to host two instances) - about $300-$500 a year. A lead tech maintainer with the skills and availability to manage development contributions from the FOSS community (10hrs/wk); a system administrator to keep the technical stack up to date (100hrs initial setup + 2hrs/week); several data stewards enforcing community standards (there is more or less infinite work there, but 10hrs/week is the threshold where data quality will not get worse from new contributions; anything above that will improve the quality of existing data); and a board to make strategic decisions; I reckon this board must not be limited to CMAA members (for instance, I am not a CMAA member and I have no reason to be one, and nor is Olivier).
Honestly, my first idea was to create an ad-hoc non-profit, or simply an informal group, because Gregobase needs more volunteers than it needs funding, and the main advantage of joining a large non-profit is the ability to piggyback off fundraising, and most of the main actors of Gregobase right now are non-Americans, nor do they use CMAA services regularly.
As I outlined in the thread, only the infrastructure and governance parts require that we move from a "single man" model to a structured organization. The development and data parts, and a bit of the infrastructure (frequent backups) can be addressed by moving from a "single man" model to a "happy few" model where a group of people who trust each other manage Gregobase. This is why I am not rushed.