| Subject: |
Business Reasons to rewrite, was: web app |
| Author: |
"Ted Roche" |
| Posted: |
2006/03/01 09:12:16 |
|
|
|
|
On 2/24/06, Dave Bernard <bernard.dave (at) comcast .D.OT net> wrote:
>
> For every person on this planet planning or executing a complete rewrite of
> a working line of business VFP system (not a developer tool) into
> .NET/J2EE/anything, I want to ask a simple question:
>
> "What were the business reason(s) for doing so?"
>
1. Scale. Client wanted to move to gigabytes of data, and their
internal programming staff and consultants they brought in could not
develop a satisfactory app. Client invested millions in DotNet and SQL
Server, and went bankrupt. Remants of the company are back to
megabytes and back to VFP.
2. Inability to find good consultants. Having gone through a dozen VFP
developers who were dBASE refugees and should not have been
developers, company went for the magic pixie dust of Java. After
millions of dollars of development (sounding familiar?), company was
bought for hundreds of millions of dollars and entire app was scrapped
in favor of the purchasing company's existing system.
(There are a lot of developers out there who write junk for code. I
don't think VFP attracts them, especially, perhaps it's just had a
longer time to accumulate them? I've seen some pretty awful stuff out
there.)
3. Painted too deeply into the corner: I supported a client for nearly
a decade who had a legacy system written by a well-known developer
early in the VFP 3.0 days. There were no best practices then, so there
were some fairly complex work-arounds. The system was very large and
very complex, and the micro-managing, penny-pinching boss would never
authorize an hour spent to rewrite something that worked, no matter
how arcanely. After a decade of making serious money out of this
application, he got caught up with a young guy who could show him
spiffy little tricks in *Delphi* of all things (out of the frying pan,
into the fire) and, not understanding the differences between
superficial GUI tricks and the deep functionality of his application,
put his existing development into maintenance mode to go on a wild
goose chase with Delphi. He was too cheap and too wily to lose his
business to this, but his best developers quit, his customer base
moved on, and when he sells out in a few years, he'll get a lot less
than he could have.
4. Slightly off-topic from the VFP re-write question, but an answer to
why not FoxPro: Corporate standards: I tried to pitch a WebConnect app
to a Very Large Insurance Company. They had standardized on: Macs on
the desktop, Novell for their network, Oracle for their database and
Netscape Enterprise for their intra-, extra- and inter-nets. I fought
this one all the way up to a one-on-one with the CIO, who tried to
explain to me that Microsoft was "going the wrong way," a view I've
come to agree with, but for different reasons and with a differnt new
direction. IT evolution slowed to a crawl in this company, and the CIO
has taken an early retirement to "pursue other interests."
5. Cost savings: Tired of paying experienced senior developers with
decades of experience in the business niche and this particular
application, PHB thought it would make sense to employ cheap VB
developers to rewrite the app in the para-dig-m of the day, VB and SQL
Server. Experienced developers moved on, weaker devs stayed on for
free training. New apps took forever to deliver, cost gazillions, and
lacked the functionality of the original. Customers wouldn't upgrade
for fewer features. Company foundered, bought up by BigCompany for
1/10th of peak worth, for customer base. Old code and new code
discarded.
In summary: incompetence, incompetence, incompetence, incompetence,
incompetence. Hmm. Guess there is a pattern <g>.
So, Dave, you were looking for GOOD business reasons to switch? I have
run into few of them:
Good apps need to be rewritten every once in a while, as cruft builds
up, and the model of the business encapsulated in the code doesn't
always evolve as fast as the business does. Software tends towards
rigidity and/or fragility. Refactoring and other advanced techniques
are designed to extend the longevity of an application, but
refactoring a gnarly old app can be more costly than rebuilding.
When rewriting, you have the glory of starting with a clean slate, and
re-examining your assumptions. New business models (software rental,
software as a service, application service provider) may be available
since the original app was conceived (probably back when we used
floppies). ACID compliance, disaster recovery, HIPAA and SOX
compliance can make new architectures a requirement. New component
models, loose coupling, multi-phase commits, heterogenous backends are
all designs to consider.
Security is a huge concern with ever-increasing connectedness,
portability and liabilities.
So, ultimately, the business decisions come down to:
1. What business(es) do you want to be in?
2. What architectures enable that?
3. What tools enable those architectures?
4. What resources do you have available to execute those designs?
When faced with a clean slate project, new languages and tools are
always a siren song. "There are no silver bullets" is a 30-year-old
quote.
However, given the specs of a couple of apps lately, I couldn't find a
justification for writing them in FoxPro. While we have a mature
language (well-debugged, well-documented and lots of support), some
great frameworks and lots of programming talent, there were concerns I
could not address: Microsoft has handed out BILLIONs in legal
settlements in the last couple of years. Security is a huge concern
and apparently something Microsoft is still not taking seriously.
Microsoft has made it clear VFP9 is the end-of-the-road for the
binaries, with some xBase decorations extending VFP9 into Sedna.
64-bit is out and support for NX bits mean some loss of functionality.
Bottom line: a single vendor who is end-of-lifing the product.
Competing languages with rich features included Perl, Python, PHP and
Ruby had no proprietary vendor lockin, no preferred data source and
the flexibility to deploy on many platforms. VFP Web deployment is a
chain of SPOFs (Single Point of Failure): W2K3, IIS, COM. In
comparison, if a mod_perl app has a problem on Apache/Linux, redeploy
on OS X or on Zeus or via CGI. Options. Choice. That's what it came
down to. The VFP solution was climbing out onto a limb with a vendor
renowned for orphaning its products.
Orphaning: In the late 80s, I sat in a room back at the Park Plaza
Hotel in Boston while Microsoft announced the rollout of the NT
platform. During the Q&A session, a fellow came up to the microphone
and explained that he was a Microsoft "partner," had subscribed to
their products and had spent years with a staff of programmers
developing an app not far from release, but targetted at OS/2. What,
he asked, was Microsoft going to do for him? His voice was unsteady,
and it was apparent that he was facing a disasterous failure. There
was an awkward silence when he finished as the crowd fell silent.
There was no noise but an occasional clink of crystal against
silverware. A Microsoftie finally managed to speak up, trying to
deflect the comment into a pitch for their new development tools. The
spell ended, but the impression remains to this day.
I can't lead another client down that path. THAT's the business reason.
--
Ted Roche
Ted Roche & Associates, LLC
http://www.tedroche.com
©2006 Ted Roche |
|
|
|
