Postman Flows: the next generation of software development
blog.postman.com
blog.postman.com
I think not. It seems like Yahoo! Pipes. It's cool but I don't see most devs using it. And FWIW Scratch is way cooler.
It could still be nice but that's exaggerating it a whole lot, I think.
Pipedream could be much further along. https://pipedream.com/
Postman somehow managed to make a slow, bloated, piece-of-crap api-tester that's somehow wildly popular. Extrapolating from that, I expect this to be incredibly successful regardless of how good it is.
To date, Postman has accepted 433,000,000 USD in funding. For a API testing tool that slowly but surely expanded in scope.
That's where things were pre-2022. Today a company of that scale asking for hundreds of millions in funding will get laughed out of the room.
All that said - know of any better alternatives?
Eventually, I figured I should try something true and tested, so I went for cURL. It takes a bit of practice and memory to learn how to do even the basics, but once it's there, it's a rock solid tool, with support for almost everything you could think of.
And then I realized curl was absolutely brilliant for my common uses because I’d write simple scripts that would mean all I needed to pass in were the arguments that changed.
I took it another step further and had my script lookup a folder that had a whole bunch of different types of requests that I could then trigger with a single command.
Finally, I created a version that plugged into fzf and made it super easy for me to run a whole bunch of common scripts trivially from the command line.
The great thing about command line based programs is the incredible ease with which you can create repeatable series of commands that can be run both interactively or programmatically, and the UX can be tailored to suit the specific needs of what you’re trying to do.
That ecosystem will do everything Postman can do, has broad support, and a much closer relationship to your actual code.
And Labview was one of the best Visual programming tools I’ve used and possibly in existence.
Still, people buy Dell Boomi and Mulesoft, so it's not like there's no market for this rubbish
But I'm not so sure they're going to get the adoption they're after.
Even the developers I know who are enthusiastic about Postman for its core offering haven't really responded to Postman's push to move more of the development cycle into the application, and I can't imagine this will be the thing to pull them in.
Edit: Unless my API framework wants to own all my shit.
They need to justify having taken hundreds of millions in funding somehow...
Plugging together APIs to services isn’t the same thing as creating software people can actually run.
Pragmatically your service Rube Goldberg machine breaks as soon as anything it’s made of makes a breaking change or goes down.
Then there’s the fact that anything built this way has zero privacy, a massive security surface area that you can’t even know let alone control, and eternal rent that must be paid for every piece that can be raised at any time.
What a dystopia we created. I say we because we all did it. We had to make software “free” and killed commercial software business models, but the only thing we accomplished was to push it all to the cloud and to a model that is substantially more closed than closed source commercial ever was.
Lesson: you get what you incentivize. Always, always, always ask “what is being incentivized here?”
This stuff has nothing to do with the majority of software development if measured by task type, even if it has everything to do with the majority of "software development" if narrowed down to browser/web-centric data presentation and editing domains.
This has nothing to offer embedded systems, creative software, operating systems, high performance computing, LLMs and so much more.
That’s a $24k/yr salary.
If you used 10 $20/mo software tools, that would cost you $2400/yr. That hardly seems unreasonable.
1. not performant
2. clunky ui makes hard to iterate - you frequently need to duplicate tabs
3. gets very messy pretty quickly
4. enables dumb people design terrible solutions (ok for small biz, terrible for big complex firms)
5. little ways to test or debug
6. underlying representation is impossible to use in git
7. impossible to refactor
The worst part Salesforce sells this extremely hard and tons of smart important people fall for it.
It's nice to see a modern version of it, though.
[0] reactflow.dev
that's a bit of a stretch
Compare to Unix. Single way to run something, one simple return status, one basic input stream, two basic output streams, standard key=value pairs that can be inherited and passed on, generic arguments, there's usually a man page that fully explains how to use it, and they can be infinitely chained.
APIs require way too much "integration" and provide too little utility, and zero composeability. They're extra work with less usefulness. But, hey, "it's the web", and we're all too chickenshit to buck any trend that starts. We're slaves to whatever shitty paradigm takes over on "the web".
The web killed computer science innovation. We're just slowly and poorly rebuilding an operating system on top of a browser. But with more complexity and less utility.