The root cause is academia still didn't adjust the rewards to take into account the computer age. People only get career points by publishing papers, not source code. Software is a fundamental aspect of most current research. We need to recognize researchers that dedicate to write/maintain the required software, awarding the fair amount of career points. Make good software documentation be the equivalent to papers for them, career-wise. It surely would improve the state of academic software. Today students still commonly have to struggle with software stuck in the 70s, usability-wise.
It would also greatly increase the diffusion into industry of techniques. I often read papers that describe an interesting approach to something I am interested in doing. But often the paper requires me to read many references and spend dozens of hours or more to reach an implementation myself. I simply can't invest that time on something that might work.
But if the code were available... then you could quickly test out if something is worth pursuing.
Arguably, by the time they implemented, they tested, at that point you're just consuming.
So another way of handling this to contribute advancement at scale is "exercise left to the reader", and then we readers should publish too.
R&D in big tech should contribute as academia does.
We have our "Chief Scientists" and "Principal Engineers" titles that recognize, essentially, a tenure track, that should be on the merits of publishing code implementing new techniques.
This is a good idea. Imagine when future Nobel prizes are given to physicists who write and maintain software just as much to those who publish papers.
This problem doesn't need any nobels to be solved. Just calibrate the rewards so programmer-scientists don't risk career stagnation if they take the time to develop and maintain tools of critical importance for science.
But imagine if physicists were as recognized for software development as much as for experimental work or theorizing. That's a much more important reward.
> Without ongoing development, FORM will get less and less usable — only able to interact with older computer code, and not aligned with how today’s students learn to program. Experienced users will stick with it, but younger researchers will adopt alternative computer algebra programs like Mathematica that are more user-friendly but orders of magnitude slower. In practice, many of these physicists will decide that certain problems are off-limits — too difficult to handle. So particle physics will stall, with only a few people able to work on the hardest calculations.
This seems very pessimistic view. FORM was written by mortal humans (with much cruder tools!) in the first place, there is no reason to believe it couldn't be done again if need be and even improved in the process; its not some irreplaceable artifact descended from heavens.
On the other hand, I imagine it would take bit more than just maintenance to make the 30 year old C codebase welcoming for new physicists to work on.
The hard part isn’t writing the code but making sure it’s valid against all the data in (for example) the PDG Review of Particle Physics book.
A bug in the code would throw doubt on every paper published with it so the expected value for any individual rewriting it or making any modifications is extremely negative for their career.
> FORM was written by mortal humans (with much cruder tools!) in the first place, there is no reason to believe it couldn't be done again if need be and even improved in the process
Sure, but how long will that take and how robust will the result be? Given the description of the design, it basically sounds like it does efficient block-level storage management in order to handle the petabytes of data that it might need to churn through, so basically you're reimplementing a DB storage engine, which is no small feat to get right and performant.
Furthermore, is there a test suite to validate that the FORM2.0 yields the same results for all past papers?
These problems are not unsolvable but they're not trivial either.
I'd rather not, thanks. Unless you meant in the sense of "take it out back and shoot it", in which case we're bloody trying already, it's not that easy.
I'm not at all worried that particle physics will slow down (...heh) because of this. It sounds like the one guy has been doing an incredible job of service maintenance over the years, sure, but these folks write software all the time.
I have a hunch that FORM is not made up of easy to grok software designs or it would probably be true that other folks would have contributed in a bigger way regardless of academic stimuli. If that is indeed the case, as is common in research-academia software product, then tie a bow on it and send it off to the Smithsonian.
I would be really quite surprised if some enterprising 19 year old whiz-bang genius from Iceland didn't re-write it in Mathematica.
> I would be really quite surprised if some enterprising 19 year old whiz-bang genius from Iceland didn't re-write it in Mathematica.
Still would be a net loss for science, as a critical piece of software infrastructure would become tied to Wolfram's proprietary black box. Also I think academic software benefits less often from a random precocious outsider fixing things, because in addition to good programming knowledge, it requires a fair amount of subject matter expertise, what is rare to be found both in the same person.
having now poked at the source code on github, yeah this is not at all maintainable by modern standards. it's the kind of code you're unable to modify until you memorize every last line and the whole problem domain.
I’ve spent a ton of time working on numerical code written in various languages and this code looks absolutely maintainable. Given how widely it’s been ported on how many platforms that’s not particularly surprising. If you’re used to business software or stuff that deals in essentially parsing and concatenating strings maybe it’s unfamiliar what a numerical C program looks like and it’s daunting looking. But frankly I feel like I could probably understand most and contribute meaningfully with a few weeks of experimenting. The comments are fairly detailed, things seem to be well compartmentalized into functions, there’s not a huge over reliance on macros, non numerical variables tend to be reasonably named.
Universities tend to track scientists’ publication records
Yet another example of administrators imposing a one-size-fits-all rule and elevating convenience over quality. Every academic I know hates how academia works, but they balk every time I suggest networking with people outside their disciplinary silos to dethrone the administrators. It's like the smartest people in the world are stymied by the idea of a collective action problem that requires working across disciplinary boundaries. For a long time I thought I was missing something fundamental about the complexity of the issue, but after 10 years of asking it seems to literally be that individuals are afraid of getting punished and they don't have any ideas beyond supporting the occasional strike action (that will leave things more or less as they are). I don't get it.
I mean you basically just described Autism. “Wow, you guys are amazing at your Special Interest and talking with other people with the same Special Interest. Why aren’t you interested at talking with these Neurotypical bean counters about finance stuff?” Not that I’m saying every physics major is Autistic. I myself know several in my graduating class who are not.
It's pretty remarkable that despite Feynman Diagrams being invented in the 1950s, before computers and computer science, they are still state of the art so many years later. It was back at the time where physicists did most calculations by hand.
You'd have thought something more efficient would've turned out by now, since there was barely a sensible definition of efficiency back then.
Feynman diagrams are a representation of an integral so moving beyond them is a question of reformulating the mathematical basis of physics.
There is some work in that direction (look up 'Amplituhedron') but overall it's not that surprising that physicists are using computers to do the same math with much more precision than they could before rather than shaking up the fundamental mathematical tools of their field which have stood since the 1600s.
Obsolescence means something else made it obsolete, meaning another program passed it technologically or provided superior user experience so usage naturally dropped off.
It’s at risk of neglect/disrepair/falling behind. If it’s important enough and something breaks hard you better believe someone will pick it up, but not until it’s a forced issue.
Key is well used, well tested and the limit by disk not memory arrangement. I am not sure about the last bit. In 1970s they do have this issue and hence you have compiler that do multiple pass to handle limited memory. And there is no operating system to help, or there is no operating system (virtual memory, swapping, virtual machine come around that time …)
I feel like particle physics has gotten a ton of funding and there just be a lot of software engineering involved at the experimental level. I’m surprised if this software is that fundamental one or all of the funded experimental programs don’t break off some modest cash to pay people to improve and maintain it.
