Re: Should Inline be private in the private part of a package spec?

["Randy Brukardt" <randy@rrsoftware.com>]

"Vasiliy Molostov" <molostoff@gmail.com> wrote in message 
news:op.wi0fliuoka8ora@aspire.local...
> Randy Brukardt <randy@rrsoftware.com> ?????(?) ? ????? ?????? Tue, 14 Aug 
> 2012 02:16:00 +0400:
>> it's mainly intended to be used
>> in the background with very little interaction. That's not always what
>> customers want.
>
> Why not to turn/move it into xml?

Because XML is a classic solution in search of a problem?? :-)

The entire point of the file format is that it is a block-structured 
random-access file (that is, Direct_IO); the compiler only accesses the 
records it needs, not the whole file. (The whole file can get quite large if 
a lot of compilation units are involved, I've seen files with 10000 
records.) Reading the entire file, especially if it is large, would be 
prohibitive. Reading it as a *text* file, which would be 3-5 times slower 
than pure binary, would be insane.

I suppose one could use blank-padding to make each record the same size 
(thus allowing random access), and put snippets of XML in each record. But I 
don't quite see the point; one couldn't create or modify such a file with 
standard tools (which would not preserve the required blanks).

Janus/Ada was designed to run on tiny host machines (like MS-DOS was), and 
as such we don't use text for anything (simply because it would take a lot 
more code to read and write than binary records do). We don't need to worry 
about hosts without decent amounts of memory anymore, but redesigning 
everything to use text files -- which still should never be modified by a 
human -- and making it a lot slower in the process -- doesn't really make 
much sense.

                                                               Randy.