Stop getting the Javascript error wrong at 6:18 Kevlin! That Canadian address error is Javascript with a $ at the beginning, which is a perfectly legal token for a variable. There used to be a convention with jQuery selector elements to use "$" at the beginning of variables, but that doesn't seem to be the case here, so no idea why they have it. Nonetheless, PHP uses "->" for object props/methods and PHP does not have a "object.prototype.toUpperCase()" function (that would be strToUpper(string)), but Javascript does.
Peraphrasing some thoughts > Kelvin : Unit test are great, integration tests are not > Kelvin 20 mins later : Always consider things in their next bigger context That's what integration tests are.. the next bigger context.
He said no such thing. He said that reliance on only integration testing in distributed system means you miss the bugs that have nothing to do with integration which are most of the bugs you will come across. Sheesh, why does the matter of testing always provoke the "No, you wrong," comments from people who don't like to listen? It's pretty.... testing!!
Makes me think if the Rust's annoyingly pedantic error handling requirements make people write better code in Rust, which would be a good language choice for these applications. However I don't know about Rust's integration to machine learning. Also how about designing systems (in case of for example military vehicles) where in case of software failure, you still have mechanical systems to get something done even if you lose the brilliant computer power. I wonder if people used to do something like if a or b is 32 bit { a/2 + b/2 } to combat the limitations.
Kevlin Henney's talks are always engaging, memorable, and thoroughly enjoyable. He's a real UA-cam for the industry.
Failure is inventible in software, always learn from your and others mistake and deliver better software continuously to users.
ok, whatever I had planned for Friday evening can wait an hour
Seen it twice now. I may watch it again in a few months, because Kevlin says things that need saying :-)
Stop getting the Javascript error wrong at 6:18 Kevlin! That Canadian address error is Javascript with a $ at the beginning, which is a perfectly legal token for a variable. There used to be a convention with jQuery selector elements to use "$" at the beginning of variables, but that doesn't seem to be the case here, so no idea why they have it. Nonetheless, PHP uses "->" for object props/methods and PHP does not have a "object.prototype.toUpperCase()" function (that would be strToUpper(string)), but Javascript does.
Great talk. Configuration is code, so true (Is that a foreshadow of GIGO?).
Another great talk from Kevlin Henney 👍👍
You taught me a lot as usual
55:00
Q: What does an incel and MS excel have in common?
A: Mistaking something as a date
Peraphrasing some thoughts
> Kelvin : Unit test are great, integration tests are not
> Kelvin 20 mins later : Always consider things in their next bigger context
That's what integration tests are.. the next bigger context.
He said no such thing. He said that reliance on only integration testing in distributed system means you miss the bugs that have nothing to do with integration which are most of the bugs you will come across.
Sheesh, why does the matter of testing always provoke the "No, you wrong," comments from people who don't like to listen? It's pretty.... testing!!
15:30 "... integration testing is of course important ..."
How do you handle errors in the self-destruct module? And how do you test it?
With a dummy explosive. ;-)
Makes me think if the Rust's annoyingly pedantic error handling requirements make people write better code in Rust, which would be a good language choice for these applications. However I don't know about Rust's integration to machine learning.
Also how about designing systems (in case of for example military vehicles) where in case of software failure, you still have mechanical systems to get something done even if you lose the brilliant computer power.
I wonder if people used to do something like if a or b is 32 bit { a/2 + b/2 } to combat the limitations.
The guy at 53:34 was not fast enough 😅