[ale] Indian outsourcing

Jeff Hubbs hbbs at comcast.net
Wed Jan 28 14:51:01 EST 2004


On Wed, 2004-01-28 at 11:48, Geoffrey wrote:
<snip>
> "Ritesh Maniar reminds me that Hexaware has scored a Level 5 rating from 
> Carnegie Mellon's Software Engineering Institute, the highest 
> international standard a software company can achieve. The others are 
> quick to note that, of the 70 or so companies in the world that have 
> earned this designation, half are from India. Over several days, here 
> and at other companies, I hear this factoid repeated like a campaign 
> talking point.

<snip>

The problem is that the CMM SEI Level 5 designation is essentially
self-awarded.  It's typically the end result of a process in which
consultants modify pre-existing software development procedures or just
clone one of their own.  Since said consultants are well-paid, they are
disincentivized to really clean the clocks of the people who are paying
them.  They do their job, hand over the paper procedures, and the rating
is applied for and achieved.  

However, even if a software shop has reached CMM Level Million-and-Six,
there is absolutely NO GUARANTEE that:

* The procedures are written in plain, grammatically correct English and
are at least self-consistent;

* The procedures can be followed with the information given by the
development client;

* The developers and project managers have training in, knowledge of,
or, for that matter, ACCESS TO the procedures;

* The timeframe and/or logistical constraints of the project allow
enough time for the procedures to be followed;

* The developers have the use of competently installed and managed
equipment and infrastructure necessary to effectively execute the
project. 

In my experience, I have seen severe project-bombing breakdowns in ALL
these areas in  claimed Level 5 outfits and that is why I trust those
certifications about as far as I can throw them.  .  

There is nothing inherently wrong with the CMM level system, but it's
primarily an academic exercise and it is best used as a guideline for
evaluating how things REALLY OPERATE in a software shop.  In my research
into the subject, I found that there is a Level 0 and even a putative
Level -1 whose characteristics far better matched my observations.

This is why I favor lightweight methodologies that treat the developers
less like the machines they develop on and more like people who are
trying to accomplish something.  

- Jeff



More information about the Ale mailing list