What is the "pedestrian" approach to self-driving cars? You mean walking rather than driving?
Re:Safari tries to fill username
by captainmuon on Saturday May 29, 2021 @06:05PM
(#27328604)
Haha pun not intended. Yeah I mean something like cruise control + lane keeping which has a limited scope but is predictable and doesn't tempt people to move to the back seat and get killed. Or maybe virtual "rails".
Re:Safari tries to fill username
by musingsole on Friday May 28, 2021 @09:43AM
(#27315219)
Rail or line following might be a more "pedestrian" approach to self-driving cars.
Ah I see, yeah, to be honest I don't get why we aren't simply doing that already. Adaptive cruise control we already have, draw one more line in the center of highway lanes, and have some warning system for 'accident up ahead, drive manually', and we could drive autonomously for any large distance without magic and without needing to have constant attention to take control when needed. This could easily have been implemented decades ago, but we didn't/don't and now we're still not at that level. So yeah, agree with the person I was replying to about the preference of non-magic there.
Re:Safari tries to fill username
by alisonkisk on Friday May 28, 2021 @05:59AM
(#27313516)
Safari doesn't write the HTML, and web page authors don't write explicit clean markup, so the pedestrian approach doesn't work in the real world.
I wonder what would happen if compilers/interpreters/lexers of various programming/scripting languages had this attitude.
Why do web page authors get a pass? I mean, most web page authors today at least also do JavaScript, where a single out-of-place character can cause the whole thing to simply break.
Yeah yeah, legacy and all that, but that's why we have doctypes.
The entire history of web browsers, from just about day two, demonstrates the folly of this approach.
The strictest, most standards-compliant browser in the world dies a quick death, every time. The most widely-used and popular browser play fast and loose with everything. If they're powerful enough, they retroactively get their fast-and-loose playing recognized as a standard.
Re:Safari tries to fill username
by madeofpalk on Friday May 28, 2021 @07:55AM
(#27314285)
> Why do web page authors get a pass?
It's not that web developers get a pass, but that browsers want to add complete features that help its users. They're dealing with the reality of the long tail of websites that exist and trying to make their users happy.
Re:Safari tries to fill username
by knorthfield on Friday May 28, 2021 @05:49AM
(#27313435)
In my opinion Safari is way too aggressive with filling usernames and passwords. Anything in any way similarly named like these it forces an autofill. And of course it overrides autocomplete="off". We definitely need an attribute that implies, "I actually know what I'm doing Safari! Please really don't autofill this."
Conversely, there are way too many websites that (intentionally or not) prevent password managers to function, which results in these kind of heuristics being implemented by password managers in order to be generally useful.
There are often two ways you can do something. In this case:
- Explicit clean markup and a deterministic GUI or
- Tons of heuristics and a magic GUI that works great most of the time but fails in ways that are hard to understand.
I feel you get this tradeoff a lot in "clever" systems. Whether it is just finding the main text on a page, blocking ads, doing search, or even self-driving cars: I often prefer the "pedestrian" approach over the "magic" approach. Even if it is a bit less powerful, it is also less surprizing.