[RobertBradbury]

Hello, I've been using Gentoo for ~1.5 years and have passed the newbie stage. Am currently knee deep into trying to solve a bug in Firefox/ Seamonkey/ Epiphany involving GDK/GLIB window mis(?)-usage that has been around for 3+ years (so I have reasonable technical understanding).

I am trying to find the best forum (wiki?) for discussing some serious Gentoo/Linux kernel problems related primarily to "interactive workstation use". I have a Gentoo Linux (mostly up-to-date w/ lots of extra ebuilds) that I am tuning for scientific (bioinformatics) use. It also functions as a low activity web, ssh, ftp, ssh server (in spite of attempts by my ISP (Verizon) to prevent this .

The primary problems that I have is that there appears to be an inability to "tune" or even "optimize" Linux for optimal workstation use. Optimal workstation use means (a) the active window in front of my face gets *top* priority; (b) alerts (e.g. instant messages); and (c) low overhead tasks (playing music, spooling to the USB printer) get attention as needed; and (d) server activities (Apache responds to incoming HTTP requests in a timely fashion); and (e) all the extra CPU time gets used effectively without interfering with (a-d).

Most Linux systems appear to be tuned for "server" applications -- which means an equal distribution of CPU and disk resources and some attempt at "reasonable" paging activity.

Here are complaints:
1) "Nice" does *not* work to achieve (a-d). A long running "emerge world" or a firefox build (which can take anywhere from 5 minutes to hours of 80+% CPU usage) even when the emerge/build are run with nice -19. This is a scheduler and/or memory management (swapping) problem.

2) The CPU and/or memory management do not handle (a) and (c) correctly. If one is running close to 100% physical memory use (in 1.5 GB of memory one can easily get there with several instances of Firefox and/or Epiphany running for days opening hundreds of tabs). Running a firefox ld will commonly consume ~500MB (I think they scale memory usage to physical memory). The problem is that when one switches to a browser page which has not been used in a while one waits and waits and waits with little CPU activity and little disk I/O activity. I would expect both to go to ~100% as it attempts to page in pages that may have been swapped out over time. It does *not* do that.

3) It would appear that high CPU activity (even when niced) produces significant negative impacts on playing music (mplayer output will cease for seconds) or printing (USB printing can stop entirely). This appears to be a complex interaction between CPU resource allocation and perhaps paging scheduling but it seems to be more CPU related.

I would like to know the best forums in which these complaints should be submitted (I looked at Kerneltrap.org but that doesn't seem to be the proper forum). Is this the right place?

Thank you,
Robert Bradbury