Using Go for Scalable Operating System Analytics

https://blog.kolide.com/using-go-for-scalable-operating-system-analytics-cb170d85b1c5

by marpaia • 9 years ago

107 10 9 years ago

blog.kolide.com

Interesting article but I am not sure about the recommendation of writing your own command parsing instead of using something like mow.cli[1] which seems fairly lightweight and self-contained.

In terms of logging I am also wondering if something like zerolog[2] might be a better choice than go-kit's logger given its higher performance

[1] https://github.com/jawher/mow.cli [2] https://github.com/rs/zerolog

marpaia• 9 years ago

I use cobra sometimes, it's not that cobra or mow is too heavy, but I think that the oklog-style command parsing pattern is no more code or boiler-plate than using a library (for many use-cases) and it reduces developer overhead because most people are already familiar with how `flag` works. The whole pattern is really just the normal `flag` library and a switch statement, but it works really nicely because `flag` is really great and has a lot of features.

justinsaccount2• 9 years ago

Yeah.. I like cobra, but it seems to somewhat force you into using a lot of globals to accomplish things.

The only downside to the oklog-style seems to be that you can't introspect the full cli tool at runtime, which rules out things like outputting documentation or shell completion files.

carterschonwald• 9 years ago

There’s at least a few factual issues in the article. The big one I noticed is that it attributed the update framework to Docker. Which isn’t true at least historically.

zalmoxes2• 9 years ago

Yes that's true. But Docker did popularize it, at least that's how I initially read about it.

It's exciting to see that TUF is now a CNCF project.

Using Go for Scalable Operating System Analytics

Related Stories