71–80 of 267 posts
This reminds me that I've acclimatised to one specific test in the PHP interpreter test suite always failing on OS X. I should go and make it either not run on OS X, or modify it so it will actually pass. That way, when someone eventually actually breaks that function, they'll notice.
Earlier quoted context omitted.
This is my thinking also... it came to me over a long time, and typically has a lot to do with why I switch jobs. At a certain point i give up struggling against the 'WTF' and just start going along with the flow, then one day i look at the stuff i am doing, realize working at the place is making me a worse professional, and move along... However I do want to note that in general these things do have real concrete ne…
> However I do want to note that in general these things do have real concrete negative impacts on the bottomline of the organization. Not true. Sometimes, things that have concrete negative impacts can BENEFIT organizations. U.S. prisons are just one of many such examples of this. A concrete positive impact would be lower incarceration and re-offending rates. American prisons have the exact opposite effect. This lea…
The author detailed precisely why I left a former Y-Combinator company, Return Path. "As far as I can tell, what happens at these companies is that they started by concentrating almost totally on product growth. That’s completely and totally reasonable, because companies are worth approximately zero when they’re founded; they don’t bother with things that protect them from losses, like good ops practices or actually…
Did you give them a ton of advance warning when you left? http://www.onlyonceblog.com/2013/09/how-to-quit-your-job
It ended up being three weeks before we agreed that I'd successfully handed off everything.
No bad feelings either way -- I'd joined the company because they had said that they wanted to 'grow up,' but that was a feeling percolating up from below. The lower levels of the company wanted to grow up and stop firefighting all the time. The top levels of the company would fight you every last way.
> There’s the company with a reputation for having great engineering practices that had 2 9s of reliability last time I checked, for reasons that are entirely predictable from their engineering practices. This is the second thing in a row that’s basically anonymous because multiple companies find it to be normal. Multiple companies find practices that lead to 2 9s of reliability to be completely and totally normal. I…
Is the implication that 2 9s are really bad, mediocre, or some other negative judgment in this context? I'm familiar with a lot of places that have trouble with even 1 9 of reliability with lots of capital but terrible culture despite employing lots of operational controls and some best practices that have trouble with one 9, so even 2 9s sounds like a dream sometimes and the way that Dan Luu worded this article make…
> well, we have some tweaks that didn’t make it into the paper. Every single time I've tried to implement a newish, reasonably complicated algorithm from a paper and contacted the authors when I've run into trouble, this is the reply I've gotten. How is it not normal? It's research after all, and if you've worked in research you should have a good idea how the paper mill works.
It may seem normal to you and people with a "good idea how the paper mill works", but it is absolutely insane from the point of view of a lot (hopefully most...) people outside of that bubble, who would likely mostly expect the results in a paper to at least be possible to replicate with the information in the paper.
This post seems so good and so self-evidently true that I'm surprised at the amount of pushback it's getting here. Not sure what else to say about it. Well, I'll say this- the "@flaky" thing is pretty mind-blowing. In my own company I have noticed many engineers have a disturbing level of comfort with deciding something is a "mystery". There are no mysteries in what we do. The test fails because something is fucked u…
Yeah...I mostly have experience only running test suites for popular libraries and for my own software...but the existence of this library is a massive WTF...such that had I read its official introductory post, I would have thought it to be really good satire: https://www.box.com/blog/introducing-flaky-a-nose-test-plugi... Can someone describe a real life production scenario in which this flaky behavior is desirable?…
But what do you do with the test in the meantime? If you disable it you lose coverage.
The best temporary state might be to re-run the test until it passes. If it fails after allowed re-runs, you have a regression.
Two 9s of availability? Half the customers I've had would be ecstatic to have even ONE 9 of availability. And those guys hardly ever ship any code due to how encumbered developers typically are in those places and release maybe once every 6 months to a year perhaps. In fact, this is basically my typical experience with most enterprise customers I've worked with as a consultant - they're unable to execute almost anything materially important and customers put up with them because nobody else is in that niche enterprise market that's keeping people employed by lack of choice / market consolidation (healthcare.gov is just a visible example - plenty more projects are even worse with perhaps even larger budgets with zero media attention).
This post seems so good and so self-evidently true that I'm surprised at the amount of pushback it's getting here. Not sure what else to say about it. Well, I'll say this- the "@flaky" thing is pretty mind-blowing. In my own company I have noticed many engineers have a disturbing level of comfort with deciding something is a "mystery". There are no mysteries in what we do. The test fails because something is fucked u…
There are plenty of times I've seen people think of things as "well, it just does X" because of a lack of depth of knowledge, but you also can't necessarily go down every rabbit hole to the bottom, unfortunately.
Earlier quoted context omitted.
The unspoken implication here is that 99% reliability is considered bad. This may not be clear if coming from a different field where 99% sounds pretty good.
Right. 99% is an A+ in school and 3 days of downtime per year.
And I could easily double it (as in half the unavailable time) if any ok ISP become available at my place.
so many of these problems have the same root cause: we don't have an effective data driven reputation system. we use gameable heuristics to track social capial. when metrics for evaluation are flawed, people behave in ways that exploit the flaws even if they increase the likelihood of failure. "we are not rewarded for necessary grunt work as much as shiny advances", for example. That's a failure of the reputation sys…
> The Chinese government is building an omnipotent "social credit" system that is meant to rate each citizen's trustworthiness.