Alex Weisberger

Working on database observability. Anything to achieve quality software.

[Alex Weisberger @amw_zero]

I’m always looking for the best way to communicate why focusing on branch coverage is wrong. Here’s a take on that using the Collatz conjecture as the code under test:

https://t.co/lxtLQ7tm9f

[Alex Weisberger @amw_zero]

would literally everything benefit from using io_uring?

[Alex Weisberger @amw_zero]

Formal Methods Only Solve Half My Problems has had a tremendous impact on me. https://t.co/CJzd5royg6.

My main area of interest has been correctness, but performance is more or less equally important. And spot on, unfortunately the tools for reasoning about both are separate.

[Alex Weisberger @amw_zero]

@paulg It seems like Replit has a gigantic product identity problem. What does this have to do with its core business (shareable repls)?

[Dominik Tornow @DominikTornow]

Why the trouble of Deterministic Simulation Testing?! Isn't classical testing enough?!

Often classical tests run sequentially in the absence of failure. Systems run concurrently in the presence of failure

DST makes testing in real conditions manageable

https://t.co/VCo6RdrH5n

[Alex Weisberger @amw_zero]

First thought: unit tests, esp. ones with a lot of mocking, would never catch an issue like this. Next one: understanding interdependent queues is the key to reliability in the presence of feedback loops. Last one: even a friggin CDN can cause an incident

https://t.co/zPk93yRjHZ

[Alex Weisberger @amw_zero]

watching this Tigerbeetle presentation, and Joran uses the phrase "the 4 primary colors of computer science." Defined as: storage, network, memory, compute.

I _love_ this framing.

https://t.co/ciDND3rsPZ

[Alex Weisberger @amw_zero]

@lauriewired You can prove a behavioral property of a system. We do this all the time. See seL4.

You misunderstand Rice’s theorem. It means we can’t prove properties generically. But it says nothing about proving them specifically.

[Alex Weisberger @amw_zero]

@ifesdjeen fyi there's a typo on https://t.co/lpTfbAV94H. Postgres is spelled "PostgeSQL" in the In-Place Update Storage Engines section.

[Alex Weisberger @amw_zero]

@DominikTornow I am curious, how did the DST catch this, since presumably you’re not running PG in the simulation since it’s not deterministic?

Or are you connecting to a real
PG instance?

[Joe @alpha_convert]

Fearlessly sending cold emails is an academic superpower.

You can literally just ask people things! And most of the time they're really excited to respond!

[Alex Weisberger @amw_zero]

@_trish_07 I bet it will still be C and Java.

[Alex Weisberger @amw_zero]

@scheminglunatic It's quite easy to write it with a for loop that uses a stack. Recursion just takes advantage of implicitly having a stack for memory.

[Alex Weisberger @amw_zero]

@penberg @DomainerRyan Instead of an arcane git command that can overwrite history, what if we instead broke each commit up into individual PRs and then merged those separately? Same effect right? Minus the clobbering of your own history.

[Alex Weisberger @amw_zero]

@TheEduardoRFS SKI is also relatively elegant, is it not? I'm not sure what you're point is.

Re beta reduction, can you elaborate on what's confusing about it?

[Alex Weisberger @amw_zero]

@penberg It doesn't destroy history. It makes history a set of atomic commits instead of a bunch of work-in-progress commits. That's assuming your PRs are atomic changes, which they are, right?

[Alex Weisberger @amw_zero]

Uh oh. I started playing around with SimPy and now I can’t stop modeling queues:

https://t.co/nWimMmyODg

[Alex Weisberger @amw_zero]

@eatonphil Performance modeling with deterministic event simulation for sureeee.