On Sat, Jul 25, 2015 at 09:06:22AM -0700, Rich Shepard wrote:
>   This started on firefox a few weeks ago. Many Web sites have a link with a
> mailing address on a page. When the cursor hovers on the link the firefox
> status line shows 'mailto...'. That's OK. But, if I click on the link
> associated with that mailto address the display is locked and not responsive
> to mouse droppings. The only response is to ctrl-alt-backspace which kills
> Xfce4 so I must again run starx to invoke the GUI.

I don't click "mailtos", instead I cut and paste the addresses into
ssh/xterm window running Mutt on another machine.  Emailing from
browsers seems like an exploit waiting to happen.  So, I don't
encounter the problem Rich has, but I see quite a few others.

One of my annoyances is that a combination of Google Maps,
Firefox, a 2048x1536 screen, and who knows how many other quirky
dependencies causes my desktop to crash to the login screen at
random times.  I must always remember to save everything else in
all the virtuals (a file being text-edited, for example) before
I use Google Maps.  Or use something besides Google Maps or 
Firefox.

While I'm sure the pathology could be debugged with some effort,
the question is *WHY* my desktop (gnome 2 or mate) is even
/capable/ of crashing (or in Rich's case, freezing XFCE) in
response to /anything/ an application can do.  This seems like
Very Bad Design, poor encapsulation, whatever.  

Firefox - somehow - has toxic hooks into the desktop environment.
How could /anything/ Firefox emits change the run state of the
desktop?  Why would the desktop respond to /any/ application
(not just firefox) with a freeze or a crash?  Don't the APIs(?)
have filters and bounds checks?

As I continue my search for a low-animation desktop that
/Does ONLY What I Tell It To With Clicks/ and /Does Not Permit
Pathologies/, I will gather stories like this.  Sometimes,
it seems like GUI programmers are intentionally malign,
adding features that behave like a keyboard that eats hands. 
If badly-written, non-regression-tested software can have
such results, what can malign software do?

As a chip designer, I feel some responsibility.  Processors
should have parallel processes that do bounds checks and byte
counting.  Moving chunks of data should be atomic operations
at the hardware level, with process-tied semaphores and flags. 
Yes, during software development, programmers would be pestered
with hardware that barfs when they make mistakes, but that is
better than /me/ barfing when their mistakes get shipped. 

Keith

-- 
Keith Lofstrom          [email protected]
_______________________________________________
PLUG mailing list
[email protected]
http://lists.pdxlinux.org/mailman/listinfo/plug

Reply via email to