Re: Support Existing Content

On May 2, 2007, at 01:37, Philip & Le Khanh wrote:

> Maciej Stachowiak wrote:
>> On May 1, 2007, at 8:19 AM, Gareth Hay wrote:
>>>
>>> To address the point I do understand, I agree, no one is perfect,  
>>> but are you really suggest something akin to using a C compiler  
>>> to take some pseudo code input and attempt to produce executable  
>>> code, by hook or crook?
>> A C compiler runs on the developer's machine. This is different  
>> from content produced by one party and consumed by many others  
>> using a variety of tools, where there is more benefit to being  
>> lenient in what you accept.
>
> That is exactly the situation that obtains with "C" programs.  One
> person writes code, others attempt to compile it using their
> compiler, their libraries, their operating system ...  A pretty
> good parallel to the way HTML is produced and consumed, in fact.

On Unix-like systems GCC holds very notable market share. GCC also  
has particular command line options and its own extensions to the C  
language.

You'll find that vendors of competing compilers (IBM, Intel, Sun and  
Metrowerks off the top of my head) have to support GNUisms in order  
to compete with GCC. The extent to which each competing compiler  
supports GNUisms depends on the customer demand for particular  
GNUisms--that is, which features are actually used by real code bases  
and build systems.

You'll also find that Linux (a pretty significant and famous code  
base) is not written in ISO C but relies on GNUisms.

And yes, sometimes it is quite legitimate to deviate from ISO C. If  
you need to use a vector processing unit and ISO doesn't cover how to  
write explicitly vectorized code and the compiler isn't smart enough  
to vectorize your traditional code automatically, chances are that  
you'll value compiler-specific vector extensions over ISO purity.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Wednesday, 2 May 2007 10:31:52 UTC