[phobos] escaping FILE's buffering tyranny

[s50]

I heard that C's stdio is mostly unchanged from the initial form which
was provided
at Unix v7. For the example of non-blocking I/O, it was introduced in
later than v7.
So "UNIX Network Programming, Vol1" said mixing them is a recipe for disaster.

I don't know why perl-guys developed PerlIO,  but I think there were
probably such
dark corners. It seems that also python 3.x has a new I/O system which
doesn't use
C's stdio. I'm searching its background to indulge my trivial curiosity now.

2010/9/30 Andrei Alexandrescu <andrei at erdani.com>:
> On 9/29/10 17:45 PDT, Jonathan M Davis wrote:
>>
>> On Wednesday, September 29, 2010 17:36:34 s50 wrote:
>>>
>>> Oops, you already have mentioned about the lookahead things.
>>> Sorry I overlooked it.
>>
>> Did you look too far ahead? ;)
>>
>> It's interesting to hear about Ruby's situation with c stdio though.
>
> I don't know about Ruby, but my experience with C++ and Perl suggests that
> there are always some dark corners in any language's file I/O, inevitably
> revolving around buffering and interaction with C. My speculation is that
> all languages must implement their stuff on top of C's stdio, which has an
> inadequate API. My big question is how come stdio has evolved so little and
> so slowly; the file API is central to Unix. But then all programs that do
> serious I/O go away from FILE* and use ioctl() et al.
>
>
> Andrei
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
>