httpd-apreq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joe Schaefer <joe+gm...@sunstarsys.com>
Subject Re: Trying to test current apreq2 CVS
Date Fri, 02 Jul 2004 22:32:09 GMT
Stas Bekman <stas@stason.org> writes:

[...]

> > (file buckets don't buffer file IO, and they always perform a seek
> > before every file-read operation).

[...]

> APR::PerlIO gives you an access to the filehandle. It can't know
> what's kind of buckets (if at all) are used behind the apr_file_t
> front, if that's what you were suggesting. 

It's the always-seek-before-read behavior of file buckets
(and which would mean a per make sense for APR::PerlIO) which
makes them so nice.  You can have multiple brigades that
contain different file buckets *which are all on the same 
apr_file_t*, but a read on any of those buckets will not 
corrupt the data in any other bucket.  This is the sort
of semantics we've always been missing (no, dup(2) doesn't
achieve this because duped file descriptors always share 
the same seek pointer).

Sorry I don't have time to say more about this now, I'm
off to the Marlins game.  Suffice it to say that I've
already written a tiny TIEHANDLE API for bucket brigades
that passes tests 6 and 7 of request.t.  I'll commit it
tomorrow after I've had time to whittle it down a bit.

-- 
Joe Schaefer


Mime
View raw message