Logo

LSSw Sustainability Town Hall - Shared screen with speaker view
Mike Heroux
16:13
Google Doc for questions: https://docs.google.com/document/d/1v1Tw6DYk4YJhZ8jZoWcSnjvzmKsk36gQugucYszPzYE/edit
David Bernholdt
36:28
C+++?
Todd Gamblin
36:40
@David that is called rust
David Rogers
37:30
D-lang?
Todd Gamblin
37:44
If only anyone used D
Ann Almgren
38:15
Was there ever a C+ (between C and C++)?
Dossay Oryspayev
38:34
Haskell/F#?
Todd Gamblin
38:47
No — ++ is the increment operator
Dossay Oryspayev
39:05
Delphi -> C#?
Dossay Oryspayev
39:31
it's sharp
Anthony Castaldo
39:33
C<< (much less code)
Mike Heroux
53:51
https://docs.google.com/document/d/1v1Tw6DYk4YJhZ8jZoWcSnjvzmKsk36gQugucYszPzYE/edit
Dossay Oryspayev
55:12
From object-oriented to aspect-oriented prog.
Axel Huebl (Berkeley Lab)
55:54
I think my question was covered by Todd's talk already
James Willenbring
56:02
Interoperability of the tools.
Hui Zhou
57:07
With modular tools, we require certain expertise and skills from users
Fernanda Foertter (NextSilicon)
57:28
Interoperability makes sense
Dossay Oryspayev
57:35
interoperabilty needs protocols
Dossay Oryspayev
59:24
linux plumbers conference
James Willenbring
01:00:38
Sustainability of the software ecosystem is a different consideration than sustainability of any given piece. An ecosystem can survive if some of its parts do not.
Andy Gallo
01:00:42
(I’m multitasking with Spack on Perlmutter right now… we love Spack… but…)
Todd Gamblin
01:02:25
I’m trying to think of other package manager projects for which there is a standard and coming up short. Conda has a competing implementation called mamba now, so I guess there is that.
Todd Gamblin
01:02:32
(Mamba is way faster btw)
Todd Gamblin
01:03:01
For folks interested in package manager standards though, we are organizing this: https://packaging-con.org/
Todd Gamblin
01:04:20
I think to enable this we’d need to separate the packages from the core tool and enable others to write one or the other more easily.
Axel Huebl (Berkeley Lab)
01:20:22
me
David Rogers
01:20:38
there are a few projects that test on minGW
Elena Pourmal
01:21:00
HDF5 works on Windows
George Bosilca
01:21:06
parsec too, mingw and native
Carol Woodward
01:21:20
Meeting followers where they are at requires investing in the activity of working with the followers, but this is typically not funded, nor publishable, nor leads to career advancement. How do we make this more "accepted" and rewarded?
Elena Pourmal
01:21:28
and we still support Solaris
Andy Gallo
01:21:59
E4S Docker on Mac has been useful.
Fernanda Foertter (NextSilicon)
01:22:04
I can’t believe solaris is still being used
Elena Pourmal
01:23:09
We have old station that is still working. Great system to expose portability problems
Fernanda Foertter (NextSilicon)
01:23:31
A testament to Solaris builders tbh
Andy Gallo
01:23:34
+1 on the security impediment to CI/CD
James Willenbring
01:24:08
I don’t think Docker is “cheating”
Axel Huebl (Berkeley Lab)
01:24:14
I uncovered tons of bugs in our Linux code bases by compiling also on macOS & Windows. Segfaults manifest in fabolous ways on different platforms
Richard Mills
01:24:22
PETSc has a significant number of Windows users, but the way people want to build PETSc under Windows is very fragmented (Cygwin, minGW, WSL2) and it's hard to support all the models people want to use.
Paul Kent (ORNL)
01:24:46
what is the difference between hpc and aws/azure/google though? The cloud providers are publicly content with their security.
Elena Pourmal
01:24:55
We are getting help from our HDF5 community
Kenneth Raffenetti
01:25:54
+1 cloud HPC is going to be critical to support
Fernanda Foertter (NextSilicon)
01:26:34
Yay stickers!
Fernanda Foertter (NextSilicon)
01:27:05
We use it here too Todd :D
Todd Gamblin
01:27:30
Yeah actually on that I think DOE should issue blanket approval for project sticker orders
Todd Gamblin
01:27:42
Thanks fernanda! 🙂
Benjamin Sims
01:27:44
Re: the comment about understanding user needs, one issue is that user studies as a field generally focuses on non-technical end users, rather than specialized scientific user needs. So there is a gap there that us social scientists could be thinking more about how to fill.
Fernanda Foertter (NextSilicon)
01:29:22
Can’t publish “we removed 1400 lines of code”
David Rogers
01:32:44
There seem to be a lot of out-of-proportion incentives with respect to other areas of human activity. Industries try to spend more money on developing user bases than developing products. Scientists (and computer scientists) will take a pay cut to do great (computer) science. Research projects have huge application areas that are mostly out of scope of what funders can target. Sustainability is at odds with adding new features.
Brian Van Essen
01:34:37
But there has to be funding on the software side to engage with these application users that are starting up with the new piece of software
James Willenbring
01:34:47
What is APS?
Todd Munson
01:35:17
Advanced Photon Source
David Rogers
01:35:20
https://www.aps.anl.gov/
Paul Kent (ORNL)
01:35:26
APS = advanced photo source. x-ray facility with lots of beamlines
Dena Vigil
01:36:30
Thank you, Mike. Great conversations.
Axel Huebl (Berkeley Lab)
01:37:01
Thank you, got to leave, too.
Todd Gamblin
01:37:46
Thanks!
Elena Pourmal
01:37:57
Thanks a lot!