[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Top 10 for old farts

David Fanning wrote:
> I suppose it is inevitable, as IDL grows ever larger, that
> we begin to pay for add-ons. RSI has already taken this
> course with DataMiner and the Wavelet Toolkit. But I am
> dead set against this proposal, Joe.

I am one hundred percent with David on this. I have the experience of
using products when once you are about done with what you are doing, it
beeps - "err, an add-on module is missing. Call for latest pricing." Not good.
Then again, take it one step further, and have a list of checkboxes when
ordering IDL: I'll buy Strings, Floats; I need no Unsigned longs; I need
For loop and Where but no While, and Plots but no T3D keyword. Sounds
strange, doesn't it? The core of IDL needs to be intact. Applications
written in IDL, like Noesys and Rivertools, can be sold separately by
those who write them, and I guess they are more suitable and
user-friendly for ad-hoc ideas verification, with no coding needed.

> In fact, I haven't written a program for a client in the
> past year that hasn't included at least one object,
> and sometimes it's easier to write the whole thing as
> an object.

I have not been using objects a whole lot until a year ago. Now I am
wishing that I did. Upgradeability of code is a lot better when it is
object oriented. Adding new functions and even totally new functionality
to the existing code is a snap.

> I've frankly pretty much given up the idea of writing
> an object book because (1) it is so damn hard to write
> a book, and (2) after going to all that trouble I thought
> only about a dozen people would buy it. (And I will hear
> from all 12 today, probably, pleading with me to reconsider,
> so desperate is the need for decent documentation.)

Effort like that may not happen to be all wasted, as RSI will probably
ask to buy the copyright and use the book in the IDL help :-)

> Someday, inevitably,
> you are going to be working with objects.

I agree. And, from personal experience, I'd say that using object code
speeds up building applications *a lot*. Even debugging is easier with
more structured arrangement of object methods than with standart widget code.