Zope Documentation (was RE: Python Popularity: Questions and Comments)
djmitchell
djmitchell at optushome.com.au
Sun Dec 30 07:08:15 EST 2001
Bill Tate wrote:
> I understand the "zen" thing, but it's not something that business
> people are going to care about when they making a decision about
> whether to use it or not.
Yep, exactly.
My last boss was a "who cares what it is? If it's open source and you
techos recommend it, we'll give it a try" kind of guy. If we could present
a justification in terms of risk/benefit that looked good, he gave us his
blessing. I switched him from thinking development=C++/Java to Perl and
Python, we replaced Solaris and AIX boxes with Linux, and we saved our
employers enormous quantities of money. When I tried to talk him into
looking at Zope, I found there was no frame of reference I could use - no
comparisons with J2EE or n-tier COM, which are probably about the closest
alternatives in terms of development approach If he was to let me take on
some work using Zope, his problems were:
- trying to convince his superiors that this unheard-of Zope thing was
better in some vaguely-defined way than J2EE and large COM solutions
- how to estimate the design/dev times using Zope. At least with J2EE and
COM there's some body of evidence out there as to how long projects take
from conception to production. We'd proven that Python compared to C++
gave something around 10:1 reduction in dev times, but Zope isn't Python
- how was Zope going to perform on hardware from vendor X. We already had
reasonable expectations as to how J2EE performed on specific hardware, and
could budget our hardware purchases with some degree of confidence. The
Zope web site suggests "it runs fast", but there's no data at all that I
could find to back it up
I knew I was on a loser even as I was saying something like "it's a better
Java than Java" as I was describing Zope's platform independence. When I
had to resort to crap platitudes like this, even though I believe it, I
knew nobody holding development purse strings was going to let me finish
the spiel.
Zope seems to have an "it's better than the rest; just take our word for
it" mentality attached to it, and solid comparisons with other tools or
performance data just doesn't appear to be out there. Sun took this tack
with Java in the early days, and MS did with COM/MTS a few years ago, and
large corporates had a belief that Sun and MS wouldn't let their customers
down (else they'd lose market share and shareholders would get upset).
With Zope not having a major corporation behind it, and the NASDAQ collapse
in 2000, there's that much less blind faith available for Zope. Even a
small pilot or proof-of-concept implementation costs serious dollars, and
that type of money just isn't around any more.
I'd LOVE to use Zope for something grander than my personal site holding my
resume and pictures of my kids, but if I couldn't convince one of the most
receptive managers I've ever had, it seems like an impossible task. It's
sort of like the conceptual leap that was required to embrace OO
development several years ago; there was a large body of evidence as to how
non-OO development worked in "best practice" scenarios, lots of project
managers skilled in non-OO development and a general feeling of comfort
that the old ways worked. Why change to OO, and take on all the unknown
development risks?
Based on the available documentation, Zope seems to need this sort of
"death or glory" approach, and hardly anyone seems willing to take these
risks at present. Although I'd love to see the next Yahoo! or Amazon
startup use Zope, there's just no way the men holding the chequebooks would
allow it.
If anyone's got any suggestions as to how Zope can be made into less of a
perceived business risk, I'd love to help make it happen, but it seems
beyond me at present.
More information about the Python-list
mailing list