OPEN: NEWUNIT Specifier

[andrew_4619]

I minor quibble, I just had a look at the help topic "OPEN: NEWUNIT Specifier". I do not use this feature but thought it was a good idea to start....

I was instantly put off by the 'green text' used for non-standard Fortran but then had the thought that I was convinced this was in the F2008 standard, which having just checked indeed it is (9.5.6.12 NEWUNIT= specifier in the OPEN statement).

Have I got this one wrong? If not a small mod to the help is in order....

[Steven_L_Intel1]

In the current documentation, black text is for features standard in Fortran 2003. In the future, we're going to drop that distinction and reserve the green text for features not in any supported Fortran standard.

[andrew_4619]

Ah OK,  I suspected that might be the case, but green always puts me off using features.... green=bad!

[Craig_Dedo]

I strongly disagree that green = bad.  Rather, green = "non-standard".  That also means green = "may not be portable across all platforms".  Notice the emphasis on the word, "may".  Many Fortran compilers support a large subset of popular extensions.

Perhaps, as an intermediate step, Intel may wish to adopt a convention that IBM used in the early days of Fortran 90.  Soon after I joined the Fortran Standards Technical Committee in February 1994, the IBM representative shipped complimentary copies of the documentation for the IBM Fortran 90 compiler for AIX to every committee member.  That documentation set used black type for Fortran 77 features, red type for Fortran 90 features, and blue type for extensions (F95 was still in production then).  Similarly, perhaps Intel would be willing to use red type for Fortran 2008 features.

[Craig_Dedo]

Here is one other thought.  I have never been reluctant to use extensions to standard Fortran.  Of course, if there are both standard and non-standard ways to do something, I will use the standard feature unless the standard method requires a significantly more convoluted way of doing something.  My choice of what extensions to use is largely determined by what platforms I want my programs to run on.  The features that I use need to be supported by all of the platforms I am targeting.  Of course, if there are only two platforms I am interested in, then I have a large subset of features to choose from.  If only one platform, then all available features are eligible for use.

[andrew_4619]

There are so many extensions that can be easily avoided using standard Fortran with the post 77 standards. Life is too short to be considering what extension may or may not be a available on another implementation or platform. 

In the case of totally non-standard platform specific stuff (like the windows SDK) I tend to lump that stuff in Fortran friendly wrappers or group chunks of code together in subroutines that perform specific platform specific tasks, It doesn't make it portable but the task of porting becomes much simpler. Whilst I only write for Windows these days who can tell what the future holds....

So is green=bad? well no not really that was tongue in cheek, but green does cause me to pause for reflection and seek a practical standard way.