Twitter | Search | |
Colin Percival
Computer Scientist, FreeBSD Security Officer Emeritus, AWS Hero, and author of Tarsnap.
18,346
Tweets
38
Following
13,225
Followers
Tweets
Colin Percival 4h
How I spent my evening: Turns out that overwriting the efiboot0 partition makes systems unhappy...
Reply Retweet Like
Colin Percival 6h
Replying to @__steele
I haven't tried yet. Busy with other things...
Reply Retweet Like
Colin Percival 13h
Replying to @_msw_
That's a great reason, but most people don't have such forward-thinking employers.
Reply Retweet Like
Colin Percival 16h
The amd64 version of the FreeBSD 12 listing in the AWS Marketplace has now been updated to 12.2-RELEASE:
Reply Retweet Like
Colin Percival Oct 27
Replying to @patio11
A potentially more value-creating topic: You've told developers to negotiate salary, but while "ask for more" may be directionally useful it's not necessarily obvious *how much* more. Perhaps some "if you look like this, you can expect to earn about this much" examples?
Reply Retweet Like
Colin Percival Oct 27
Replying to @patio11
It's probably not the most *valuable* writing you do, but I find that some of your most *interesting* writing falls into the category of "a Western geek's perspectives on Japan".
Reply Retweet Like
Colin Percival Oct 27
It had occurred to me to wonder if the Marketplace team had been instructed to prioritize product listings which run on Graviton! Wouldn't make so much sense for cases like this were it's one product with two listings, but this is something of an edge case for the Marketplace.
Reply Retweet Like
Colin Percival Oct 27
Replying to @awsmarketplace
For some reason while the was able to get the arm64 listing () updated in time for the release announcement, the amd64 listing () is still stuck at 12.1-RELEASE. Hopefully they'll deal with that update soon.
Reply Retweet Like
Colin Percival Oct 27
FreeBSD 12.2-RELEASE is now available. EC2 AMIs are available for both amd64 and arm64 platforms.
Reply Retweet Like
Colin Percival Oct 26
Funnily enough, back in 2006 I did spend a while estimating how much interest I would get. Even then the answer was "not much".
Reply Retweet Like
Colin Percival Oct 26
I'd say that I'm happy to keep on collecting picodollars in exchange for keeping accounts around for years and years... but honestly I didn't know was using until just recently, so this falls into the category of "not deliberate". Self-serve SaaS is awesome.
Reply Retweet Like
Colin Percival Oct 25
It's the same with code analysis tools -- most of the complaints I see from Coverity or the Clang analyzer are false positives, but the combination of a tool flagging something and me thinking about the code which was flagged produces useful results.
Reply Retweet Like
Colin Percival Oct 25
I understand. That's why the output of the statistical analysis should be "here's something you might want to look into" rather than "here's the new formula you should use".
Reply Retweet Like
Colin Percival Oct 25
Replying to @_msw_ @car_mcnaught
Both humans and models designed by humans are liable to have blind spots and systematic biases -- but they probably won't be *identical* so comparing the two will at least let you identify some of the problems which need to be fixed.
Reply Retweet Like
Colin Percival Oct 25
Replying to @_msw_ @car_mcnaught
I think the easiest option is to develop statistical models which will say e.g. "hey humans, you seem to be overvaluing Ivy League degrees and undervaluing non Ivy League degrees... might want to look into that".
Reply Retweet Like
Colin Percival Oct 25
Replying to @_msw_ @car_mcnaught
Right, by ML I mean "regression model which identifies correlations for humans to examine". Not a state of the art black box!
Reply Retweet Like
Colin Percival Oct 25
Replying to @_msw_ @car_mcnaught
Interesting. I kind of expected that there would be a routine "after 2 years, review work and decide whether / at what level employee should have been hired" which would feed back to a giant ML system which determines which hiring inputs are under/over valued.
Reply Retweet Like
Colin Percival Oct 25
Replying to @car_mcnaught
You didn't really answer my second question. Of course Amazon does research and refines processes; that's table stakes for a large company. What I'm asking is whether that research includes "these are people we shouldn't have hired" as an input.
Reply Retweet Like
Colin Percival Oct 25
Replying to @car_mcnaught
Do bar raisers get a veto on hires, or does their feedback simply get added into the pool? Does Amazon have any "in hindsight, we shouldn't have hired this person; where did the interview process go wrong" loop?
Reply Retweet Like
Colin Percival Oct 24
Given the reference to blank ballots, I'm guessing they're getting ready for arguments about whether misspellings and other errors are close enough to allow a ballot to count for a candidate.
Reply Retweet Like