On 17/01/07, Nadav Har'El <[EMAIL PROTECTED]> wrote:
On Tue, Jan 16, 2007, Oded Arbel wrote about "Re: Why are GNOME applications (and applets) take so much [EMAIL PROTECTED] memory ?": > I usually see a problem after about a week of usage - after a reboot it > behaves itself for a few days. I rebooted this morning, and now Evo is > down to 400MB. With the mad race to get better and better hardware, it seems that people forgot how to write efficient software. One of the lost arts is preventing memory leaks.
Keep an eye for Andy S. Tannenbaum's keynote at LCA this morning on http://lca2007.linux.org.au/Programme#head-6af3ad9cefbbb05127e86c3d2f00c2542a1bb75e (I'm sure the slides/audio/video will show up later). He talks exactly about this - how his PDP-11 with 64kb RAM used to boot in 4 seconds while his top-of-the-line Xeon server takes two minutes to boot, or how a VAX shared among 80 users and with 1 Mb of ram gave good service to all of them. He quoted someone from Microsoft to the effect of "Software gets slower faster than hardware gets faster" Ah, and another good quote - "Software is gas - it grows to use all of its container". (He then goes on to a different path - explaining why Minix 3's tiny amount of kernel-mode code is important in a world of bloated software). knows what). What I can't understand is when the memory of such a program
grows over time, forcing you to reboot every week (like you said).
Maybe it's not the clock applet. You can't 100% deduct it from the data so far, can you? Contrast this to more "modern" software, which leaks megabytes *every day*
(if not every hour), and nobody is even trying to do anything about it... Alas...
Indeed. I remember in my degree studies, Prof. Tarsi tried to explain why it's important to write more efficient programs even as you run on faster hardware, as much as people appreciate his intelligence, the argument didn't always come across... Then again - that's why there is enough work for GOOD programmers...:) --Amos
