After a long hiatus, let’s restart the fortran-lang community call.
This session will focus on the status of the ecosystem: fpm, stdlib, lfortran, etc. Feel free to propose additional discussion priorities, upcoming releases, etc.
The final date will be announced here on or before the evening of Sunday, September 6. The call will be held on MS Teams (link posted here once the date is set) and recorded as usual. Propose agenda items below!
Good question - That’s exactly the kind of thing the call is for: figuring out where each piece of the ecosystem stands and who’s driving it.
Adding it to the agenda - I’ve never been involved in the registry development myself, but it would be great if someone who worked on it, or wants to pick it up, could give a short status update during the call. Or post one here in the thread, and I’ll share it during the call.
Hello all, a few quick items to discuss during today’s call (link above, 17:00 CEST / 8:00 AM PDT). Bring anything else you’d like to raise to the open floor.
fpm update (Federico): v0.13 features & profiles - conditional compilation, per-compiler/OS flags, optional dependencies. (@hkvzjal, others) discussion on the registry status/implementation. What’s next? LFortran support status; PR backlog.
Ecosystem update: maintainer effort, funding status, opportunity for contracts and grants.
Free AI/compute credits for open source: several providers run OSS credit programs - worth applying for fortran-lang projects?
AI policy for fortran-lang projects: do we accept AI-assisted PRs, and with what disclosure? How do we handle low-quality AI-generated PRs/issues? AI in docs and review triage. Proposal: a short written policy for the org.
Open floor and next steps: owners, dates, next call.
Yes, the meeting was recorded - I’ll post it on the fortran-lang YouTube channel (getting access sorted) and link it here, together with notes and next steps in the next day or two.
Thanks @FedericoPerini for organising the call and, with @certik , for sharing a recording here (which I have just finished watching) and to those who participated in the call. I’d like to respond to a few items.
One of the items discussed in the call was funding, particularly in relation to fpm. I’m leading on the FortranFuture project, in the UK, which is funded by a UK research council. FortranFuture includes flexible funding to support others, though mostly this will need to be ‘ring-fenced’ to the UK. More generally, I wonder whether there might be opportunities for some of the flexible funding to ‘seed fund’ other opportunities, perhaps fpm but other activities too (of which one I mention below). Happy to discuss sometime with people who are interested.
Another item discussed was AI. Broadly speaking, my sense was that the discussion covered two areas: 1) policies and good practice around using AI in one’s Fortran development, e.g., in relation to further developing fpm; and 2) the behaviour of AIs (LLMs) and how to ‘control’ that behaviour. People may be interested to know that item 1) was discussed in various sessions at the UK’s RSECon26 recently. Slides are on their Zenodo space and videos are on their YouTube space. Then, for items 1) and 2), people might be interested to watch James Kermode’s talk on the use of Agentic AI in scientific computing which includes some discussion of Fortran. A post is on LinkedIn, with slides and materials on GitHub . James talks about patterns and anti-patterns when using agentic AI and provides advice, experience and worked examples too. Without prejudging anything, I wonder whether FortranFuture could use flexible funding to extend/deepen the work he has done to provide advice in relation to items 1) and 2) for Fortran. Again, happy to discuss with those who are interested.
It was interesting to hear too about continuing innovations, e.g., @Beliavsky 's work on a Python to Fortran transpiler, serving as an exemplar of how Fortran can continue to contribute. Not everything need be, or should be, a one-way street away from Fortran.
Finally, I agree with closing comments about the value of establishing a recurring meeting.
Yes, and even if the target language is C, Fortran can play a role, since the mapping from NumPy to Fortran has been done by Pyccel and my transpiler, and since there is a modern Fortran to C transpiler (which I need to improve).
. I have marked a few dates as possible.
