Appending data to pre-exisiting files

[Emil Block]

My comments were not meant to be critical, and I agree with  your 
reasoning for
adhering to the standard. I was only suggesting that it would be nice to 
have the
extension available. 

I have been compiling Fortran code for about 20 years with about as many
difference Fortran compilers, and can only recall one time where
'access=append' was a problem.  G77, Sun FORTRAN,  COMPAQ/INTEL Fortran
compilers currently include the extension.

I've 'watched' the hard work the people developing Gfortran have put 
forth, and
have nothing but admiration for their efforts.

The only minor problems I have encountered  as a Gfortran user are due 
to removing
 features in G77 that do not adhere to  the new standard, but could be 
included as an
 extension -- not a big deal in any event.

By the way -- I like your Mark Twain quote , and I don't paint fences.

Thanks for your comments.

blime

Richard E Maine wrote:

> On Jul 27, 2005, at 1:59 PM, Emil Block wrote:
>
>> g77 supports  'access=append' , and gfortran should, in my view, 
>> support this
>> non standard form as do the many fortran compilers that I've used 
>> over the years.
>
>
> I have personally used many compilers that do not support that particular
> nonstandard form. In fact, I doubt that most of the compilers I use today
> support it. If you have code that depends on this, you'll have porting
> problems with other compilers, even if you haven't seen them yet.
>
> I have also long regarded this as a particularly ill-advised
> form in that it confuses two different concepts (access method
> and file position), thus inviting difficulties relating to future 
> enhancements.
>
> For example, it doesn't "fit" very well with f2003 stream access. It 
> is quite
> reasonable to want to append to stream files, but this form doesn't
> accommodate that because the "access" keyword has been used for
> something that isn't an access method, leaving no way to specify
> the actual access method. The access='append' form just assumes
> that sequential access is implied. The standard's form of
> access='stream',position='append' works just fine.
>