The Old Joel on Software Forum: Part 3 (of 5)

Grudgingly admitting defeat

Have you ever had to implement a technology you *hated* in a proof of concept? Did you do your best to make it work anyway? Did you have to eat crow and admit it'll do the job (possibly dooming yourself to working with it forever)?

Or did you, either intentionally or by simply not trying hard enough, "allow" the technology to fall flat on its face? [I'd recommend withholding your name on this one]

(This thread brought to you by the "allowing the tech to fail" observation by Portabella in the VLDB thread)

Philo

Philo
Friday, September 26, 2003

Not to hijack your thread, but I wonder if anyone has ever done a proof of concept which didn't ... prove ... the concept (i.e., the initial analysis lead you down the primrose path to unmaintainability and unscalability (is that a word?)).

To answer your question, Philo, no, I've never had that experience, cause I'm always right. =-)

Alyosha`
Friday, September 26, 2003

I have done a proof of concept that showed we didn't have the knowhow to do the whole project. The development team swore if they put a little more time in the prototype, it would work. The performance was an order of magnitude from what we needed so we punted and ended up licensing something instead.

To Philo's original thread, nope. Usually self fufilling prophesies.

pdq
Friday, September 26, 2003

Well, I've been "forced" by senior management and sales&marketing to implement something I was convinced and adamant just couldn't work.
More than once it's turned out I was wrong; when I really tried I made it work... I've always chosen to regard this as conclusive evidence of my own genius <ahem>

SteveM
Friday, September 26, 2003

> something I was convinced and adamant just couldn't work.

Well, the longer I stay in IT, the less I am inclined to say "Can't work", because I've seen so many amazingly bone-headed ideas actually made to work.

I call this the follow-through coefficient, and claim that it can be described by an equation like this one:

worth of idea * follow-through coefficient

In other words, even really good ideas will fail if they do not have enough oomph, and even really bad ones can often be made to work, given sufficient push.

Please note that just about *anything* can be made to work if we have control over what the requirements are, since we just change the requirements to match what we actually did. I've actually seen projects which were run this way, though usually not commercial ones.

Portabella
Friday, September 26, 2003

>>proof of concept which didn't ... prove ... the concept

yes. I would say about 95% (or higher) of my research is thrownout.

Tom Vu
Friday, September 26, 2003

Sometimes, even on a commercial product, it simply makes sense to tweak the requirements to match reality. We oftentimes create a UI design that turns out to be not *quite* do-able. Instead of spending three days trying to wrestle with Windows Forms to get it to look *just perfect*, we just change the requirements.

Note, however, that we only do this for those requirements that really don't matter, such as small tweaks to the UI. For important requirements, we would cut the feature before we'd work the requirements around the difficulties of the particular technology, since going forward without doing it right would drag down the quality of the product.

Dave
Friday, September 26, 2003

What's wrong Philo? They picked Oracle, didn't they? They probably did it just to spite you if you've been talking up SQL Server to them as much as you have here.

Wedge
Friday, September 26, 2003

I've had a few. My attempts to use ActiveX from C++ have all ended in scrapped code. Pretty much anything I've done with Visual Basic has wound up scrapped, since it was obvious that it was unmaintainable in the long term.

My proof of concept for getting rid of embedded HTML in ASP programs went right out the window. Actually, I followed that proof of concept a few months later, and have since used a variant of it in almost all of my software.

Clay Dowling
Sunday, September 28, 2003

That's one of my pet peeves: embedded HTML or not.

Leonardo Herrera
Monday, September 29, 2003

Tom Vu,

95%, wow! What type of field do you work in to allow such a failure rate?

Ged Byrne
Tuesday, September 30, 2003

Is this Camel related?

Just me (Sir to you)
Tuesday, September 30, 2003

LOL! Nope.
Prompted by my own tendency to resist tech I don't like/trust/know, and the effort I have to put into making fair evaluations.

To be honest, there's nothing similar on my plate at the moment - the thread was prompted by the previous thread I mentioned.

On the Camel front, I understand they lost another senior architect, the third one pretty much stays in his cube and watches everything go by, and one other person has taken ownership of the DB schema to the degree he doesn't allow anyone else to review it.

I notice tomorrow is Oct 1st, and I suspect they won't be shipping 1.0...

Philo

Philo
Tuesday, September 30, 2003