[uClibc]uClibc-based server

Nick Jennings nkj at namodn.com
Tue Dec 10 20:47:15 UTC 2002


Hi Tony,

 I've actually been experimenting with uClibc myself for the past month or
 two, for similar reasons. However the more I learn about it, and deal with
 it, the more I'm finding it's not going to be a perfect as I thought. 

 As far as it being lightweight, yes, it's good for low memory systems, such
 as embedded devices. However, in this article:

 http://www-106.ibm.com/developerworks/linux/library/l-lwl1/?t=gr,lnxw12=LLP1

 There is mention that it's focus is on size, and not speed, and there is
 often a trade-off for small size. I'm not sure how it measures against
 glibc in terms of speed (which was specifically why I wanted to use it
 in the first-place), and I'd be interested in any data which can shed
 light onto the subject.

 So, if you want your mini-distribution to run in extreme low memory 
 situations, I'm sure uClibc will excel. However, if memory is not so much
 an issue as speed, I'm not sure that uClibc is going to give you much
 bang for your buck (factoring in the blood, sweat, and tears it takes
 to get things working). 

 Hope this helps.

- Nick Jennings

On Tue, Dec 10, 2002 at 11:13:28AM -0500, Tony Tonchev wrote:
> Hi Guys,
> 
> Looking for some advice...
> 
> I'm new to uClibc and fairly new to Linux in general. I also realize that there is a steep learning curve ahead. I've spent the last couple of months compiling a mini i386 Linux distribution against uClubc. I know that uClibc isn't intended as such, but my purpose is a compact server distribution with versions of Samba, Apache, PHP, MySQL, etc. without the bloat of glibc. That, coupled with a nice shiny web-based admin interface sounded so good to me that I had myself a project.
> 
> I have a bootable/working root with all essential utilities such as lilo, busybox, tinylogin, cfdisk, ex2fs utils, ncurses, nano, etc. The last week or so was spent battling libnsl. I spent time modifying config.h files, looked at disabling nis/yp support, etc. The magnitude of required work seems to grow with every tweak I make. I'm at the point of throwing in the towel...
> 
> The idea was that the minimalistic approach should result in high performance on mid/lower end machines. I like the idea of cramming as much useful code into the CPU cache as possible. Since uClibc is compact, the result, I thought will be positive.
> 
> If I'm wrong in that assumption and in general headed the wrong path, someone please correct me so I don't dig myself any deeper then I already am ;-)
> 
> Any advice on the subject is welcome...
> 
> Thanks...
> 
> Tony



More information about the uClibc mailing list