Linux's approaching Achilles' heal

On Sat, 17 Nov 2007 06:02:02 -0800 (PST),
Evenbit <nbaker2328.DeleteThis@charter.net> wrote:

> On Nov 16, 11:00 pm, Keith Kanios <ke....DeleteThis@kanios.net> wrote:
> <<snipped>>
>> I don't see how "straw man" applies here. I am simply commenting
>> from the appreciation of being a system-level programmer.
> It is obvious that if you are indeed a "system-level programmer" who
> is worth his salt, then you would have _some_ understanding about
> modern memory management issues (it is clear from your responses that
> you do not). When we issue a call to an OS asking for a chunck of
> memory, the OS responds by looking for an area of _contiguous_ free
> memory space of the size that we request. So, you see, it is
> perfectly possible that an attempt to allocate 50Gigs will fail, while
> subsequent calls to the same OS function asking for 10 instances of
> 10Gigs each will succeed.

That's odd... I was under the impression we had this thing called
paging, on modern operating systems. This has two effects; one,
applications are actually allocated memory in complete pages, and
secondly, those pages can reside anywhere in physical ram, and they'll
still appear contiguous to the application.

The only time this might be an issue, is with DMA, where a component
external to the processor (and hence without the benefit of the kernels
page tables) needs to access data across two or more pages.

Mind you, I'm not a systems level programmer either...


>> If one process is hogging all of the physical and swap memory, other
>> processes are being deprived of that memory. Ask Windows users how
>> appreciative it would be to lose one application's worth the data
>> instead of losing all of your data due to the entire system becoming
>> unresponsive.
> Wouldn't the better choice be to not lose ANY data??? Why do Linux
> developers consistantly shoot for standards that are _below_ that of
> Windows developers? Why should end-users tolerate a less-stable
> experience -- especially when Linux-fans consistantly "bill" Linux as
> the better(TM) product??

You, mate, are an ass. Every time I have run out of memory on a
Windoze system, the entire system crashed. My wife who still uses
Windoze will attest to that. All current unsaved data, in all
applications, gets flushed down the drain when not even Ctrl-Alt-Del
will respond, and you have to reach for the power button (because
modern machines don't come with a reset button anymore).

Every time I run out of memory in a Linux system, one application gets
hosed, _usually_ the right one. Though occasionally it's like my GUI
panel or something, which subsequently gets re-started, causing
something else to die instead, and occasionally it'll roll through two
or three unlucky minor apps before it hits the right one. It can also
be a bitch when it's the X-server itself that it decides to kill, but
such is life. I just sit back and watch for a few minutes, after which
I have a system that's at least stable enough to save down anything
that has survived, and either restart the X server myself, or give the
while system a thorough cleanout with a nice soft restart.

It's still a damn sight better than the Windoze way of just locking up
the entire frigging machine, and hosing everything indiscriminately.


>> If the problem is actually with running out of process (virtual)
>> memory, then I can think of more graceful ways to handle such
>> out-of- memory situations.
> This is indeed the issue at hand -- being "more graceful" than killing
> the calling application and preventing any error reports from being
> issued.

The question, is how exactly do you do that, without allocating
additional memory?

Come to think of it, how do you figure out when enough memory is really
enough? My system will quite happily (albeit a little slowly) run with
3-4 times the base memory allocated, as long as no single application
accounts for twice the base memory. In Windoze, it starts to die well
before that.


>> ... instead of Java, C# and garbage collection wiping incompetent
>> asses? It would be appreciated, but highly unrealistic when software
>> is market driven. Quality is no longer a factor, it is just reduced
>> down to time and price.
> This is the very mind-set and attitude which will get Linux labelled a
> "has been" in the OS history books.

But that is the mind-set that exists industry-wide. One only has to
look at Microsoft's business applications, most of which palm off HTTP
and XML as gods gift to software developers. They've rammed their
stock-standard HTTP/XML libraries into places they simply don't fit,
and focused on making the application look pretty so end users will
like it, and not notice the utter shite under the hood. I've seen it
time and time again. Most of the good quality innovative developments
I've seen of late, have come from Linux, not Microsoft.


So I really think you've got your head on backwards, mate. Linux's
achilles heal, if anything, is that fact that it's doing the job
right, rather than cutting corners and building lock-in boxes, in an
attempt to rule the world.


Fredderic

On Sat, 17 Nov 2007 05:35:24 -0800 (PST),
Evenbit <nbaker2328 RemoveThis @charter.net> wrote:

> On Nov 16, 9:57 pm, Keith Kanios <ke... RemoveThis @kanios.net> wrote:
>> I wouldn't call audio a *critical* system.
> Audio is certainly a critical system for those users who are not
> blessed with the normal human attribute of being 'sighted'. Blind
> people do not depend on either screen graphics or text from a video
> monitor -- they are able to use a PC solely via the audio feedback.
> Why should library developers be granted exclusive permission to
> determine which systems are *critical* and which are not? Shouldn't
> these decision be left for the application programmer to decide?

They're not. Both systems get pretty much the same regard, as far as I
can see. But one would offer the suggested that without sight, there'd
likely be more memory for the audio system. Plus audio generally has a
lower memory footprint, and so short of audio editors and other
high-end music creation software, a simple screen reader is far less
likely to draw the application killers gaze, and far more likely to be
automatically restarted even if it did.

You know, I may have missed part of the thread, but it seems to me that
tugging on the accessibility string really is another step down the
ladder for you.


Fredderic