Earlier quoted context omitted.
Why don’t you just disable automatic updates?
Good question. Because for security reasons you want to stay up-to-date on software that connects to various websites. At the same time, from a functionality point of view I wished I'd never have to update anything.
Ask HN: Firefox connection problems after enabling DoH?
301–310 of 410 posts
Re: Ask HN: Firefox connection problems after enabling DoH?
#302Re: Ask HN: Firefox connection problems after enabling DoH?
#303Earlier quoted context omitted.
It is because: (1) I did not explicitly enable this (2) The telemetry setting seems to have re-enabled itself on some update (3) I don't want any services from Mozilla, I want a browser (4) It worked until it blew up revealing that in fact, I suddenly did have service dependencies And finally, the reason I use FireFox is exactly your last sentence, so to see that they are slipping this in under the radar is a pretty…
This is caused by HTTP3. The only way for you to have never experienced this issue would be if Mozilla never integrated HTTP3, which isn't going to happen because the major service providers like Google and Cloudflare are adopting it. Telemetry and updates happening to use HTTP3 merely pushed the issue to the foreground, it would have happened eventually and possibly in a way that got fixed more slowly because it was…
You argue that it wasn't an automatic update: I am pretty sure that the first install on this machine did not have HTTP3 support and that automatic updates pulled it in, end of story right there.
As for the telemetry issue: that's even worse because telemetry and all other forms of communication with the mothership other than automatic updates have been disabled on this machine, and has been silently re-enabled without my consent. That's a pretty gross violation of trust, and if that in turn causes me to lose a morning then that makes it even worse. Fortunately, today is not an interview day, but if this had happened two days ago the consequences would be terrible.
Finally, automatic updates would ideally just fix security issues and not introduce new, possibly unwanted functionality. I carefully select my tools for their purpose and I absolutely hate this brave new world where critical stuff suddenly stops working because some company could not be bothered to take their end users' interests a bit more serious.
Google and Cloudflare were not implicated here, it was FireFox that stopped working, Chrome still functioned just fine.
Re: Ask HN: Firefox connection problems after enabling DoH?
#304Earlier quoted context omitted.
>If they don't turn this around then I hope others will step up and reclaim the web! Have you any idea how complex a browser needs to be? I hope this team of plucky, idealistic coders can keep up with all the latest and greatest web developments devised by the thousands of engineers at Google et al. The web as an open, human-sized system is dead.
If you keep the browser ultra modular, I think it can work. You need to have one component for layout, one for css, one for the DOM, and you can use external libraries for media playback, javascript, networking, and so on. I think a decent layout or CSS library would be useful outside of a web browser, too. Then I would also only focus on the subset of websites that are "documents", not "apps". If I could decide, HTM…
Re: Ask HN: Firefox connection problems after enabling DoH?
#305This happened to me and a co-worker too, and I could resolve it with a `sudo killall firefox` and re-starting it. In contrast to all the heat from many of the comments here, I'll say this: shit happens to any company. My trust in Mozilla is not (yet) shaken just because a bug made it into production, even if that bug made it there due to bad decisions or even (hopefully temporary) lack of vision. I, for one, am not c…
Re: Ask HN: Firefox connection problems after enabling DoH?
#306I "fixed" it by refreshing Firefox. You can refresh Firefox from [Help] menu item, and choose [Trouble shooting Mode...]
Re: Ask HN: Firefox connection problems after enabling DoH?
#307What worked for me was disabling HTTP3 support with the 'network.http.http3.enabled' key in about:config and then restarting Firefox. Seems like it's stuck in the 'SocketThread', repeatedly doing this: 2022-01-13 08:20:53.075936 UTC - [Parent 4106991: Socket Thread]: V/nsHttp Http3Stream::OnReadSegment count=333 state=4 [this=7f6e295623a0]
Yep, works. However, you need to magically restart the browser. Apperently, it runs some background processes whose names contain neither the character sequence "firefox", nor "mozilla", according to pgrep.
Re: Ask HN: Firefox connection problems after enabling DoH?
#308Re: Ask HN: Firefox connection problems after enabling DoH?
#309Earlier quoted context omitted.
Oh weird you're right, I don't think I've ever interacted with that extra address/search bar. I wonder what they were thinking, that seems redundant and a bit silly
It is for people like me who doesn't want every domain I ever type in to be sent to Google or somewhere else for autocompletion. I mean, most of all lkve autocomplete, but when you type in . or what have you, are you sure you want everyone of these keystrokes to go to Google? If not then you want a separate search field.
We have the same feelings on this - I also would prefer an explicitly separate search bar and address bar.
Edit: or maybe I'm understanding Ekaros' comment incorrectly? Either way the thing I'm talking about is that there's a combined search/address bar always visible in FF, and when you open a fresh window or tab there's a landing page which has an additional combined search/address bar.
Re: Ask HN: Firefox connection problems after enabling DoH?
#310And here I go explaining my friends and family I helped switch to Firefox that the browser I've installed on their computers got broken at the beginning of a new day while Edge and Chrome work fine... personally, I know very well bugs happen and I'm not mad, but this certainly hurts the confidence in it for the non-tech people.
Windows 11 almost but not entirely broke HDR. It kinda-sorta works for some things, sometimes, but most apps that used to work with HDR back in Windows 10 just can't any more and are forced to use SDR with sRGB gamut only.[1]
The Win UI SDK regressed from WPF and lost all wide-gamut or HDR support at the API level (which probably explains the above). As in, Microsoft literally removed a wide swath of floating-point color support along with the wide-gamut scRGB color space. We're back to 8-bit RGB arrays in sRGB only as the only option. Like in the 1990s.
Windows Server 2022 can't activate its license unless it uses the UTC time zone. Why? Because Microsoft employees test using Azure VMs, which... use UTC by default.
Speaking of color: Firefox for a while just... stopped doing color management. Then it worked again after a few weeks when it updated.
But it might not be Firefox's fault, because Windows also seems to randomly turn color management off, or force it back to sRGB silently.
If you have a laptop with one of those hybrid Intel+NVIDIA GPU combinations, then HDR games don't work at all with the built-in display, they all report HDR support as N/A. But they will work with external displays!
Speaking of HTTP/3 in Firefox: I've had it permanently disabled because the early releases would leak about 10 GB of memory per minute and lock up the browser very quickly. But not quickly enough to prevent the automated test suites from passing.
... and so on...
The point I'm trying to get to is that in 2022 we've achieved this state of affairs where human beings don't do actual Quality Assurance any more as a job. It's all automated and those people have been summarily fired.
Any issue that is invisible to a DevOps pipeline is Not A Bug and will ship broken. If it ships in a working state, that's probably just a lucky accident, and a subsequent patch will break it for sure.
All of the issues above are caused by automated test suites one way or another. Automated tests are literally blind to output color rendering; testing that requires a physical monitor. Automated tests are almost never set up for long-term testing for things like memory leaks, because they have to run fast. Automated tests use default, vanilla settings for the host OS. Automated tests don't have funky hardware combinations. Automated tests don't measure "jank", or inconsistent performance issues, Etc, etc...
What you're experiencing is the end result of all of this. NOBODY is sitting down and validating the end-result from the perspective of a human user sitting in front of an actual device. That final quality assurance is just not there any more, and hence we're all embarrassed when we have to "show off" some piece of IT tech and find that it's just a broken mess.
/rant
[1] This bug has apparently been fixed in some beta, and might ship around the middle of this year. In other words, Microsoft is perfectly content to break display output on their consumer desktop operating system for six months and just leave it at that.