[sf-lug] installing minimal debian and building on that

[Rick Moen]

Quoting Alex Kleider (akleider at sonic.net):

> I've been following the discussion of Desktop and window manager
> options with interest and would like to try my hand at beginning
> with a minimal (terminal only) installation and then gradually
> building  a 'to suit my needs' system.

Coolness.
 
> My first hurdle is which of the following to choose:
[...]

cat /proc/cpuinfo:

[...]
> model name  : Intel(R) Core(TM)2 Duo CPU     U9400  @ 1.40GHz

x86 family, which narrows it down to IA32 (which Debian calls 'i386') 
or x86_64 (which for historical reasons Debian still calls 'amd64')
architectures.  We could look up that processor on the Internet to find
out if it is x86_64-capable, but instead we can determine this from the
CPU flags:


[...]
> flags       : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
> mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe
> syscall nx lm constant_tsc arch_perfmon pebs bts nopl aperfmperf pni
> dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1
> lahf_lm kaiser tpr_shadow vnmi flexpriority dtherm ida

The flag 'lm' stands for 'long mode', indicating that this is x86 with
64-bit extensions, i.e., x86_64, which Debian calls amd64.  So, that
would be the binary image that leverages the Intel Core 2 Duo's ability
to do 64-bit memory addressing.  But let's talk for a moment about
memory.

The main benefit of x86_64 over IA32, aside from increasing problems
from the approaching 32-bit Unix epoch
(https://en.wikipedia.org/wiki/Year_2038_problem) is the ability of
64-bit CPUs to address dramatically more RAM.  32-bit CPUs are limited
to a 4GB possible address space,  Because of tricks required such as
reserved stretches of the memory space, the effective free RAM available
to the OS on a 32-bit CPU can be a bit less, even if there is 4GB
physical RAM available.

By contrast, with a 64-bit CPU, dramatically larger total addressable
RAM is possible.  In theory, this could be 16.8 million terabytes of
memory, or 2^64 byte, but real-world motherboard implementations always
have (so far) had limits imposed elsewhere in the hardware that lower
the limit to something that's very spacious but not _that_ staggeringly
high.

https://stackoverflow.com/questions/8869563/how-much-memory-can-be-accessed-by-a-32-bit-machine

So, cutting to the chase, that means you should always use an x86_64 OS
and applications if your x86 CPU has the 'lm' flag indicating 64-bit
capability, right?

Well, sorta.  It depends.  Most cases, the answer would be yes.  But
consider a x86_64 machine that has 2GB total RAM and you're certain it
will never have more.  In that case, the theoretical ability to address
more than 4GB of RAM is never going to be used, so you might consider
for a while running an IA32 build, at least until problems from the
approaching 32-bit epoch in 2038 become prominent.  

What's the advantage of an IA32 build in that situation?  Binaries and
data require more RAM in 64-bit computing, because 64-bit pointers take
up more memory and for some other reasons.  So, it's a bit of a
tradeoff.  Other subtle aspects are covered here:
https://en.wikipedia.org/wiki/64-bit_computing#Pros_and_cons

To get the best of all worlds, have an 86_64 CPU, run an x86_64 OS, and
have lots of RAM.  Then, you won't mind the slight bloat from 64-bit
pointers, and will be able to party on January 19, 2038 like it's 2037. 


> I think the amd64 version is the one I want but perhaps it's i386?

Almost certainly the former.