[illumos-Developer] webrev: 278 get rid zfs of python and pyzfs dependencies

[Ivan Nudzik]

My comment was not offensive, just a comment saying some opinion. I must
type more emoticons. ;-) No problem with meritocracy (I remeber), just
not to fall into autocracy.

There were some misunderstanding on my side. I agree that zfs command
shouldn't have python dependency/embedding, it's also better to have
filesystem commands statically linked when things goes bad on FS. That's
for why I didn't took this combination into account and started comment
about python zfs bindings.

I know that Illumos going to be very stripped and my comment is to think
about only in case that there will be a decision to keep/add some
scripting included.

I'm not saying that 15M lines of C code has to be rewritten to Python.
Python position is the same as shell scripting is. Difference is that
not calling bunch of system commands, but directly calling
syscall/ioctl/whateverAPI of op. system. Benefits of it are clear.

It's never late to learn new language. It takes ~3days for skilled C/C++
developer to learn Python, just not to afraid. ;-) But C/C++ and Java
developers are usually religious to programming language of their
preference. ;-)

I.

On Fri, 2010-11-12 at 09:22 -0800, Garrett D'Amore wrote:
> I don't want to degenerate into a language debate, please.  This is not
> about the the best language, its about managing our *dependencies*.
> 
> Python, perl, mono, java, C++, ruby and all those other languages have
> their advocates, and they each have good and bad points.
> 
> However, I will point out some things:
> 
> a) each of these languages should be *optional*.  A distro might choose
> to add layered software on top, but we should let that be a distro
> decision rather than a decision made by the core system.
> 
> b) opinions raised by folks who aren't contributing code ... well lets
> just say they don't carry very much weight.  I'm of the opinion that
> that opinions about the underlying tech should be weighted by amount of
> contribution.  That might not be entirely fair, since a few people
> contribute a great deal, and a lot of others contribute nearly nothing
> besides their opinions.  But I did say at the beginning that illumos was
> a meritocracy, didn't I?  Let's see some merit!
> 
> c) Anyone who says Python is necessarily easier isn't looking at the
> full picture.  We have some 15M lines, the vast majority of which is in
> C.  I can code up a fix for a simple problem in 10 minutes just like
> they probably can in Python. (If its in Python, its going to take me
> *far* longer, because I first have to learn the language.)   Of cours ea
> difficult problem is going to take longer no matter which language is
> used.
> 
> There is an argument to be made that its better to make things easier
> for the "proven contributors" than to penalize them in order to try to
> lure some hypothetical future contribution. :-)
> 
> d) Python will always be required by the *build* system, mostly because
> our SCM (mercurial) depends on it.  I'm talking about removing *runtime*
> dependencies.
> 
> 	- Garrett  
> 
> On Fri, 2010-11-12 at 15:12 +0100, Ivan Nudzik wrote:
> > I see Python bindings/api to ZFS management as a great benefit. Having
> > similar ones for resource management (counterpart to newtask,
> > dispadm...), network (counterpart to dladm, ifconfig...), IPF,
> > zones... and Illumos going to be No.1 choice for many cloud/storage/HA
> > systems adopters. We are now in 21st century and there is no place for
> > C-mainly approach.... the same as we left assembler-mainly approach
> > few decades ago. 'Hard work' has to be done by C and 'the higher
> > logic' in productive scripting. It greatly improves solution delivery
> > times and stability/security too. Python class of every manageable
> > part of ON - ultimately sharp 'swiss knife' from best Unix ever! ;-) 
> > On other side I understand the effort of keeping Python interpreter
> > build solid and up to date. But comparing to Perl (Mono - there is
> > some noise about it), Python takes less time to manage build. If there
> > should be a scripting included, Python is the best choice from my
> > point of view (and skills with other scriptings).
> > And Ad. to comments in thread.... about *Real* developers and 'picking
> > 1 language': Real developers are those, who makes the most stable
> > functionality on the least lines of code - no matter which
> > combinations of langs; speaking only 1 language - it's a sign of
> > analphabet nowadays - for example Google separates Java developers
> > from "Java developers" by requiring also knowledge of coding in
> > Python.
> > 
> > That's my opinion...
> > 
> > I.
> > 
> > On Thu, Nov 11, 2010 at 9:17 PM, Garrett D'Amore <garrett at nexenta.com>
> > wrote:
> >         http://cr.illumos.org/view/jpj55p64/
> >         
> >         This work was done by another Nexentian.  The goal is to
> >         eliminate the
> >         use of Python from zfs, which was "split" into having some
> >         percentage of
> >         its code written in Python and the rest in C.  (Fixing this is
> >         a step
> >         along the road towards making python installation "optional"
> >         at Runtime.
> >         The other dependencies are beadm and IPS.  We're working on
> >         both of
> >         those as well, and not all distros will use IPS anyway.)
> >         
> >         Please provide feedback.  The timer on this expires in one
> >         week.
> >         
> >                - Garrett
> >         
> >         
> >         _______________________________________________
> >         Developer mailing list
> >         Developer at lists.illumos.org
> >         http://lists.illumos.org/m/listinfo/developer
> > 
> > _______________________________________________
> > Developer mailing list
> > Developer at lists.illumos.org
> > http://lists.illumos.org/m/listinfo/developer
> 
>