I really wanted to like Supabase, and decided to adopt it as the back end for a mobile app I'm building. So... I was invested to some extent.
But I had to abandon it after wasting weeks trying to do simple things. The biggest problem is the lack of documentation. Fundamental parts of the system are undocumented, like the User table. There's no doc on how the columns function, so I couldn't determine why a user is marked as "confirmed" (presumably through E-mail or other validation) immediately upon insertion to the table.
There's also no full documentation of client-library syntax. For example the Swift library: There are a few examples of queries, but no full documentation on how to do joins (for example).
And just try to use your own certificates; something that I've been doing for years during iPhone-app development was impossible with Supabase.
And why? Because these simple scenarios appear to be distant outliers for Supabase. It's as if nobody has ever brought them up before; and even if they have, nobody has been able to answer the first questions about them.
If you're not building a single-page Web app that just lets people browse a database, Supabase doesn't seem to envision your application.
So I went back to a plain Deno back-end, which is what I was building before trying Supabase. In the amount of time I wasted trying to scrounge up documentation and fruitlessly asking questions in forums and Discord, I was able to learn and implement authorization, and then get back to work building a product.
Maybe all this money will let the Supabase team hire some people to document their product.
Let’s hope that a tiny portion of the $200M goes towards documentation. If they spent $5k on professional writers they could get something useful. For $50k something great. And for $500k they could have an entire suite of highly produced explainer videos with great post production.
$5k won’t even get you native English language writers, $50k might get you one. One decent writer…for 6 months. Y’all really don’t know the value of skilled non software engineer professionals do you?
Why not just throw AI at it? Seems to be the best use case. So get a startup to fix this startup…so on and so forth…
Citation: professional writer with technical writing experience (also out of work)
My mom was a writer all her life. The last 5 years she's done more and more editing. After LLMs kicked off, and I mean to the month of them starting to hit headlines her work plummeted, then spiked with tons of garbage, and is now levelling off with her usual workflow.
For me, that was a strong signal that everyone gave it a go, found it too difficult to generate quality stuff, and reverted.
>Because these simple scenarios appear to be distant outliers for Supabase
You've only talked about 2 things : Lack of documentation (which I somewhat agree with) and using custom certificates. Custom certificates is not a "simple scenario" and I don't blame Supabase for not spending time on this. I fact I would prefer they work on other things (like documentation !).
It is 100% a simple scenario. I can't speak for Android, but you have to use HTTPS now for calls in an iPhone app if you want to get it approved. That means you need to deploy certificates to your test devices, simulators, and development machine.
"Lack of documentation" speaks to several apparently routine use cases being outliers; otherwise, they'd be documented. I already talked about the User table Supabase provides (and populates in unexpected ways), and about the Swift library that you have no reference for formulating joins through... another critical and expected ability.
That is the root of the problem with these batteries-included frameworks: lock-in.
Once you encounter a problem they either don't want to solve or haven't solved, your only choices are either:
- start layering on hacks (in which case you quickly get into case where no one and nothing else could help you)
- decide not to do that-thing
- do a rebuild to get rid of the batteries-included.
Personally I think something Supabase is great for toy projects that have a defined scope & no future or a very early startup that has the intention to rebuild entirely. Just my opinion though, maybe others feel more comfortable with that level of lock-in.
Even something like Heroku is miles better because they keep everything separated where your auth, database, & infrastructure aren't tightly coupled with a library.
But I had to abandon it after wasting weeks trying to do simple things. The biggest problem is the lack of documentation. Fundamental parts of the system are undocumented, like the User table. There's no doc on how the columns function, so I couldn't determine why a user is marked as "confirmed" (presumably through E-mail or other validation) immediately upon insertion to the table.
There's also no full documentation of client-library syntax. For example the Swift library: There are a few examples of queries, but no full documentation on how to do joins (for example).
And just try to use your own certificates; something that I've been doing for years during iPhone-app development was impossible with Supabase.
And why? Because these simple scenarios appear to be distant outliers for Supabase. It's as if nobody has ever brought them up before; and even if they have, nobody has been able to answer the first questions about them.
If you're not building a single-page Web app that just lets people browse a database, Supabase doesn't seem to envision your application.
So I went back to a plain Deno back-end, which is what I was building before trying Supabase. In the amount of time I wasted trying to scrounge up documentation and fruitlessly asking questions in forums and Discord, I was able to learn and implement authorization, and then get back to work building a product.
Maybe all this money will let the Supabase team hire some people to document their product.