[DUG] Embarcadero article

[John Bird]

Not sure of the pitfalls etc, but the idea is surely worth pursuing, it is 
after all pretty much the holy grail of software.  The ability to at least 
move a pure VCL over several platfrorms to give native versions on each (or 
complile all on Windows for instance) would be wonderful.

The areas where difficulties arise no doubt will be the moment the VCL or 
hte programmer wants to start calling win32 stuff, a rewrite of VCL 
internally to abstract that sounds like a terrific idea.   For the 
programmer will there be some even limited ability to call some win32 
routines - ie some subset emulated as part of the newer VCL?  Maybe maybe 
not.

Some thoughts as I use to maintain a codebase of a large number of software 
packages that could be recompiled across PDP's, DOS, Windows and SCO Unix:

    -the program code was as much as possible written to have no OS IfDefs 
in it, what remained where minimal

    -anything that was OS variant was abstracted to a library routine that 
was internally ifdef'd and recompiled once for that OS in the approprate 
version of the library.

    -Some entire programs were also largely different versions for each OS, 
as they involved generating system commands, mainly programs to do with 
backup and printing as the commands used for each were OS specific backup 
and printing commands.   These programs still shared most of their code base 
and the areas that differered were separated out to text command parameter 
files.

    -Some fixed values most use literal characters for that you did not 
think about were also abstracted into constants that were redefined on each 
platform, eg CRLF as that differs everywhere.

Considering the intracies of dealing at the low level with PDP terminal 
escape sequences, DOS ANSI character sequences for colour and screen 
positioning, and Unix termcap entries, it all became a nice uniform 
programming environment that did get most of the benefit from each OS visual 
features.

So especially if the stuff that really differs from OS to OS can be 
abstacted out at least to some degree so that the calling code is the 
(mostly) same then this has a real chance of being a winner.  I think a 
benchmark to aim for is it should be about as easy or difficult as catering 
for the existing versions of Windows (98,2000,XP,Vista) and with and without 
themes as far as the programmer goes.

This ideally would be an alternative way to virtualising an OS to get cross 
platform functionality...

Actually the more I think about it -  the more an important an idea this 
is!.    Hardly anyone else is doing this, Microsoft certainly are not. 
Other cross platform languages like Qt or XULRunner (Mozilla) do not have 
the userbase of Delphi or have their own specialised area..

Also there is the possibility of getting some serious commercial backig for 
such a project, I imagine Skype would be just one of the vendors interested 
in supporting a version of Delphi for other OS's - I don't know how they do 
the Max and Linux versions so far, but they are always behind the Windows 
versions.

I would add to the list of OS-X/Windows/linux  also look at mobile 
platforms, and if at all possible consider a version for the Web using 
Delphi code - might it be possible to run a recompiled version of a Delphi 
app as a web page????   If  you  could do that you may not need the other 
native versions of Delphi at all.....My guess however is that current Web 
technology is still a long way below native program in functionality and 
simplicity.

well Delphi for PHP is there, and ASP already does something similar using 
C# code for both program and Web pages......

But surely the future of computing is one super language to use anywhere, 
including Web pages,  not owned by any one corporation.   We may not able to 
get there yet, maybe until HTML  and Ajax is much more developed, but that 
is a goal to not take ones eyes off.

John