Out of curiosity, were the other distros that the pages displayed correctly
on the same or on different machines?


On Thu, Aug 15, 2024 at 1:49 PM Keith Lofstrom <[email protected]> wrote:

> Debian 12 seems not to error-correct corrupted animation
> files associated with html web pages.
>
> Weeks ago, I kvetched about screen tearing on Debian 12 -
> web pages displayed weirdly-patterned pixellated garbage
> with Firefox and Brave browsers.  Those pages displayed
> properly on Debian 11, and on (Much) older Ubuntu and
> Redhat distros.
>
> My "cure" emerged by accident; I noticed that the Ethernet
> switch in-line with the "hot side" (feeding the firewall
> machine) of my Ethernet connection was MUCH slower than I
> expected after my Comcan't to Ziply internet connection
> change (pro tip - if you can, move to Ziply PRONTO).
>
> Connection speed slowed MORE during a 60 second Ethernet
> speed test.  The speedometer graphic pointer, displayed
> on a hot-side after-switch connected, fast "sacrificial"
> Chromebook, dropped from 60 mbps to 40mbps.
>
> Something was warming up and getting worse, seemingly
> inside the hot-side Ethernet switch.
>
> Connecting the Chromebook (running ChromeOS) directly to
> the Ziply optical network terminal, bypassing the Ethernet
> switch (and disconnecting my house network from the 'net),
> resulted in a 250 mbps speed test.
>
> So, on a hunch, I replaced the suspect Netgear 308 Ethernet
> switch with a new-out-of-box Netgear 308 ... and I got 250
> mbps on both sides of the switch.  After that, the pixelated
> screen artifacts on web pages vanished on the Deb12 machine.
>
> Finding and fixing my problem still leaves me with a vexing
> question - why does Deb 12 display corrupted images if fed
> a corrupted data stream, when Deb 11 and other distros do
> error correction, and (more slowly) display good images?
>
> Sadly, this realization occurred AFTER I "destructively
> diagnosed" the failed crap-packet-spewing switch.
>
> In the capable hands of a Debian kernel guru, a stream of
> crap packets (with some good ones) might help debug the
> error-detection-and-correction code that FAILS in Deb 12.
>
> I fixed my corrupted-data-stream problem by replacing the
> hardware that corrupted it.  I can afford to, and afford
> to keep new-in-the-box ready spares waiting on the shelf.
>
> In a more-fragile network environment, replacing the
> hardware may not be an affordable option.  I imagine a
> rural third-world village with an intermittent data
> connection, using old hardware culled from first-world
> e-waste.  They won't have a noiseless connection and
> perfect hardware for comparison debug.
>
> They might blame Debian 12, then switch to bootleg Windoze
> instead.  Their kids don't learn Debian, we lose some
> future gurus.
>
> I should have kept the borked Ethernet switch; a noisy
> packet generator might help developers write and debug and
> test more robust code.  In the wrong hands, hell on Earth.
>
> On the other hand, a machine built and coded to produce
> INTENTIONAL and REPEATABLE corrupted packets might help
> design more robust future versions of Linux.  If "Corrupter"
> was open-source code, it might also help marginally-equipped
> people become Linux adepts and future contributors.
>
> I'm probably over-thinking this.  However, the first step
> towards more robust, data-defect-tolerant Debian 13, 14,
> etc. is more-complete automatic testing before release.
> An upgradable device designed to intentionally create
> calibrated defective data might be a lucrative product
> and test instrument for professional coders everywhere.
>
> Keith L.
> --
> Keith Lofstrom          [email protected]
>

Reply via email to