>Delphi was a GREAT product.

Maybe it was.

I had to work on Delphi (2006? I think) native applications that were hooked to an Oracle 9i database.

The language wasn't too bad, and data handling was easier than .NET by far.

Problem was Delphi itself was one of the buggiest priducts I've had the displeasure of working with since maybe VFP 3.

The develpment environment would crash for no particular reason when you were just ediitng a source file. It would nicely give you a message that it had crashed. Your work since last save was gone.

It would crash while debugging.

It would crash in the help system.

It would crash on startup.

Walk away from the machine for a while and come back - it's crashed again.

I had it on 3 different machines at the railroad where I was on contract. All would do it. I reinstalled completely on 1 of the machines (OS and all). Same result.

The language seemed nice.

I never want to deal with it again. I've had a number of calls to do so - Apparently there are some fairly large Delphi applications that can't find developers to support them. I can see why.


>
>Yes, it was. And Borland was aggressively targeting FP people to capitalize on unease about MS's intentions and try to build movement towards Delphi. They took some of us on week-long retreats with gurus. Performance was great and it all looked really easy when demonstrated. The only reason I didn't move was: data. Delphi's C/S data handling could not match the VFP Beta.
>
>Perhaps people forget that despite all the subsequent downgrading, the RV allowed you to write apps with free change handling and 100% sql-injection-proof against practically any database with an ODBC driver. VB certainly didn't have that, Delphi wasn't as good with more than a single table and Powerbuilder was just Powerbuilder. ;-)
>
>As for bugs: lets start with the grid. Yes, many of us wrote around the bugs that we all grew to know. Actually there were issues as late as VFP9 to do with cell background colors behind supposedly transparent controls. But FP2.6 didn't have a grid at all. We used to have to position browse windows over forms and jump through hoops to make them stay there and behave sensibly as users navigated around. It's not as grids suddenly took a backward step with VFP.
>
>What about long field names: whether with local or C/S databases, FP2.x required truncation of long field names and manual management of the result. Anybody who ever used FP2.x against Oracle databases with the usual long fieldnames will remember the fun. ;-) But in VFP fields retained their names and screeds of code could be replaced with a single command. I get the feeling that not many here were doing much C/S in 1995 because the improvement was monumental over what else was out there.
>
>Views did have bugs, some of which could not be worked around. There was a nasty memory leak that utilized all resources as you kept requerying. That was bad. Our workaround was to split processing into 750-transaction segments and then spawn a new instance whenever that limit was reached. FWIW at the same time Word 2 had a nasty memory leak that also used all your resources if you opened and closed lots of documents. Who holds that against Word now? We prevented it the same way.
>
>Performance with tabs: an issue, but FP2.6 didn't have tabs (though some of us had rolled our own) and the arrival of tabsets did lead to busy forms, often with hundreds of controls and grids. I think it may have been Anders Altberg whose solution I first saw. It would be called "lazy loading" today, iow not loading pages until we actually need them, to reduce instantiation baggage. That made a big difference. FWIW, lazy loading is one of the features needed for POCO in 2010.
>
>As for the rest: I must say I'm surprised. Having done well enough on the beta to be in the fortunate minority awarded a shiny new VFP box that arrived long before the official release date, I released my first 32-bit app in August 1995 the moment that Win95 was first released, which happened first in the world at midnight in New Zealand in case anybody is wondering. ;-) Apart from the memory leak and a couple of other annoying glitches, it all worked and the customer was more than satisfied to receive a 32-bit app with C/S capability. Of course I agree that VFP5 was better and of course I understand why somebody might have been willing to wait a year or two for it, just as they might have been willing to wait a year or eight for POCO in NET. ;-)