Re: LedgerSMB 1.3.6-rc1 released

["David A. Bandel" <..hidden..>]

On Thu, Nov 24, 2011 at 18:11, Chris Travers <..hidden..> wrote:

[snip]

>
> Ok.  So with a server, my primary concern is that upgrading things is
> always a somewhat risky process.  Basically, when you upgrade, you
> risk breaking things, so generally speaking you want to have longer
> terms of support and a more lazy upgrade cycle.  And there are a
> couple things about the Fedora upgrade process that are particularly
> risky.  For example yum distrosync will upgrade major versions of
> PostgreSQL for you, meaning if you didn't dump your data first, your
> PostgreSQL db is now unusable.

I hope you're joking.  While a Debian upgrade is not the easiest thing
in the world (they do their best to make it so, though), at least this
is not an issue.  You run postgresql versions in parallel until you
finally get rid of the old one because the new one works properly for
you.

I've gone through 8.4->9.0->9.1 (and many before I can't remember)
upgrades w/ no problem (personally I use testing, even for my own
servers).  With Debian, they leave both database versions running
until you've dealt with any upgrade issues.  So far, I've only had
one, that was when a contrib stored procedure I was using didn't make
it from 8.4->9.0 (but it appeared again in 9.1).

I run Debian testing because I find the constant small upgrades (vice
really big ones) is just so much easier -- upgrade daily, one or two
small packages at a time.  In fact, _my_ servers (not my clients) all
run Debian testing.

>> . .
>>
>>
[snip]
>>>
>>> For non-dev installations of LedgerSMB, in my opinion, you really
>>> need
>>> a distro with long-term support.  This means one of:
>>> 1)  RHEL and friends (CentOS, Scientific Linux, etc)
>>> 2)  Debian Stable
>>> 3)  Ubuntu LTS (and friends, like Mint LTS)
>>> 4)  Anything else with a long support cycle.

This comes down to:
1.  RHEL Enterprise (and derivatives)
2.  Debian stable (and derivatives, although Ubuntu is, I believe,
Debian testing vice stable, but with a slightly longer cycle between
updates except for security updates)
3.  Slackware (is anyone using this?)

>>>
>>> The problems that Darald brings up are real ones.  There may be
>>> advantages for us devs ignoring these and working on short-term
>>> support releases ourselves.  However I would not today use these in
>>> setting up servers for customers.
>>

As I mentioned before, I find Debian testing to be sufficiently
bleeding edge to catch problems before my clients see them -- and yes,
I eat my own dog food (a la Debian testing).  I would never suggest
Debian unstable (they really should call it Debian broken, as it
usually is).

I diverge from Debian testing on one point:
Except for _base_ Perl modules, I load and update directly from CPAN.
I'm actually not sure why distros don't just do this.  It's a lot
easier than trying to keep thousands of individual modules updated.

>>
>> Agreed.
>>
>>
>>> Debian is not a bad distro, and neither is Scientific Linux.

Scientific Linux will likely be Debian testing -- guess I should check
it out sometime.

>>
>>
>> I might have a look at a virtual SL setup now on your recommendation!
>
> Scientific Linux has a couple things to recommend it as a Server OS.
>
> 1)  It tracks Red Hat Enterprise in both support and versions.
> 2)  It includes a lot of things that RHEL doesn't include
> 3)  They beat CentOS to releases.
> 4)  They are backed by heavy Linux users (Fermi Labs, CERN)

I believe we do need devs using both RH and Debian based development
platforms (though not necessarily both), and it sounds like that's
already happening.

I have had in mind putting up a Debian stable Virtual Box image w/
LedgerSMB that users could just import and go.  Guess I'll need to
work on that ASAP.  But will also put up a Debian testing development
environment (with a few extra goodies) built from svn.  Virtual Box's
clone facility makes this task easy.

I guess I could also put up an RH-based box as well, although it is
probably better if someone more familiar w/ RH did that.  It's been a
few years since I admin'd anything RH based (so long, in fact, that
yum didn't exist).


>
> Best Wishes,
> Chris Travers
>

My .02 cents,

David A. Bandel
-- 
Two things are infinite: the universe and human stupidity; and I'm not
sure about the the universe. -- Albert Einstein
Visit my web page at: http://david.bandel.us/