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

[Rick Moen]

Quoting Alex Kleider (akleider at sonic.net):

> MemTotal:        3933692 kB

4GB physical memory, yes.

I'm actually a bit mystified why that isn't 4194304 kB (1024 * 1024 * 4), 
but maybe it's net of 260612 kB memory reserved for hardware uses, or
something like that.

> ... which I interpret to mean my machine has 4G of RAM (which has
> served me well for many years and I fore see no need to increase it-
> Indeed that may be the limit for my ThinkPad X301 machine anyway.)
> That coincides with the limit you mention for the (called by Debian)
> i386 OS.

Lenovo ThinkPad X301 is not exactly a model.  It's a model _series_.
But yes, according to
https://support.lenovo.com/us/en/solutions/pd011277, all of the X301
models appear to max out at 4GB, comprising a pair of 2GB PC3-8500 DDR3
SDRAM occupying the unit's two SODIMM sockets.

> It's unlikely that either I or my machine will live long enough for
> 2038 to be a problem!

Don't be so hasty.  It's subtle, and there's a reason why I worded my
coverage as I did, and included a link to the
https://en.wikipedia.org/wiki/Year_2038_problem article for more
information.  People have been _already_ encountering software failures 
as far back as the early 2000s because of date handling that must
correctly deal with future dates.  These will increase with time.

To conjure up a hypothetical, imagine that you're in charge of the home
loans department at your local bank.  One fine day in 2004, your
software for mortgage amortisation tables starts giving payment and
interest results that you, as an experienced banker, can see simply must
be wrong.  You cross-check on a variety of computers and using different
software, and thoe particular mortgages' all have screwy numbers.
Fortunately, the bank trust's your judgement over the computers , and
halts loan processing for a day while a specialist is called in.

The guy re-runs the calcultations on a64-bit computer running a 64-bit
Unix running the same banking software recompiled for 64-bit operation.
Now, the tables are correct.  WTF?  Oh, wait, it's 2004.  This is a
35-year mortgage.  That means that the software is needing to crunch
dates after the 32-bit Unix epoch.  The bank now hustles to deploy new
computers with new 64-bit OSes for the loan department.

Maybe you catch the consultant before he/she leaves and say 'Where else
in this business are we going to start getting mysteriously wrong
results and mysterious software failures on account of 32-bit time_t?'
He or she replies 'Dunno, man.  It'll probably come as a surprise.'  
Memo to the CTO:  'When's the next computer upgrade cycle?'

> My guess there fore is that given this additional info, you'd change
> your recommendation to i386.


I actually didn't recommend 'amd64'.  I said it's complicated.  ;->  
Again, there's a reason why I pointed you to
https://en.wikipedia.org/wiki/64-bit_computing#Pros_and_cons , because
there are nuances and trade-offs.

On balance with a 64-bit x86 machine and 4GB total RAM in 2018 running
Linux, I'm pretty sure I'd go with 'amd64'.  IA32 is already retreating to
niche, for one thing.  Many distros have dropped it entirely, by this
late date.  Running an x86_64 distro will mean that the OS will be able
to manage your 4GB RAM more intelligently.  Increasingly, fixes and
security protections are making it into x86_64 software but not IA32.
And so on.  If you suspect that a particular x86_64-compiled binary is
suffering bloat from 64-bit pointers, etc., you _can_ try instead
running on your 86_64 distro the same program compiled as IA32, to see
if it's significantly more svelte in its RAM profile.  It'll run just
fine as long as you have any required 32-bit libs to support it.

I'm not sure I'd bother with that experiment, though, because I would
guesstimate that the difference would not usually be significant, and 
that the test wouldn't be worth mytime and trouble.

Now, if you'd said that this ThinkPad had only 2GB RAM and you were dead
set against maxing out the RAM, _then_ I might advise going with an IA32
distro for now until the ongoing death of IA32 started to become a
bigger problem (or until you upgraded to something with more RAM



> PS I've never heard anyone on this list speak ill of top posting
> although it is seriously frowned upon by the people at Python Tutor.
> Does this list have an opinion?

The problem with the term 'top-posting' is that it obscures the issue by
misleading those new to the discussion, and even some who've done it
before, about what the issue _is_.   Consequently, the discussion
usually descends into meaningless digression immediately.  And that is
one of the reasons the resulting debate is widely deemed noxious.

Better framing would be to state that the issue is interleaved vs.
non-interleave replying, with the implication of trimming quoted 
text (i.e., removing everything in the prior post not being currently
and specifically responded to).  But there are a couple of problems 
_even_ with thus reframing the discussion:

1 (minor problem):  About 90% of the participants will have had very
little experience and no understanding of the underlying logic of
threaded mail readers, ones where the norm is that if you seriously 
need to read a prior post, you find it in the thread tree and look at it
there -- such that you don't _need_ to have the entire goshdarned
conversation recapped, upside-down, in every single post.

2 (major problme):  It turns out, an even higher percentage of
participants typically never, or almost never, modify the configuration
of any of their software.  They just fail to mention that, act as if
they're listening to advice about changing MUA defaults to (e.g.) do 
interleaved quoting, but ignore that information silently because they
just never consider doing any such thing.

To elaborate on the latter point:  Ever since the middle 1990s, most MUA
software aimed at novice users, and even software aimed at the
semi-clued like Mozilla Thunderbird, has come preconfigured to _not_
interleave text and to automagically quote (by default) the entire
previous thread down below the draft message in progress.  Therefore, 
the 95%+ of new users since the 1990s passively do exactly what the
software default prompts them to do, and would never do anything else
because it would require visiting and modifying their MUA's
configuration, which for them is Right Out.  

With that fact as background, the other main reason the discussion is 
generally futile emerges:  Most contributors to the (alleged) discussion
immediately rationalise whatever it is they already do.  So, while the
discussion purports to be one to decide or persuade Internet best
practices, instead it becomes a I'm-justifying-what-I'm-doing faceoff.


For whatever it's worth, I have done since about 1978 Internet-standard
interleaved quoting with trimming of all text not _immediately_ being
replied to.  It is part of my habit of (among other things) attempting
to model desirable Internet social conduct and good netiquette.

I could vent my (generally scathing) opinion of the other thing, but
that would generally be futile because of the background facts I
mentioned, so I'm not going to.

(Quibble:  Mailing lists don't have opinions.)