httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dirk-Willem van Gulik <di...@covalent.net>
Subject Re: Filtering issues.
Date Thu, 13 Jul 2000 02:09:33 GMT


On Wed, 12 Jul 2000, Greg Stein wrote:

> On Wed, Jul 12, 2000 at 03:59:18PM -0700, rbb@covalent.net wrote:
> >...
> > This is fixable, but only by using read/write buckets (which imply a
> > memcopy) with the ap_r* functions.  Most content generators will need to
> > be re-written to take full advantage of the optimizations available to
> > filters, but they will work with filters regardless.
> > 
> > This problem will crop up sooner or later in the patch submitted by
> > Greg.
> 
> No. It won't.
> 
> My filter design considers the 95% case: no memcpy() is needed *unless*
> somebody desides to "set aside" the output.
> 
> Conversely: you designed for the 5% solution and force the other 95% to
> conform to that solution. You must memcpy() data into arbitrary-lifetime
> buffers.

Actually, one of the point's of 2.0 is that we can stop worrying about
this. This was _THE_ number one premisse for the 2.0 work. So this is
a non issue IMHO. Unless we suddenly see 2.0 as a 1.3 on steroids. 

Secondly, if you look at what some of the legacy modules do; pass pointers
to static buffers which they later modify, it should be clear that in
'legacy' mode you _always_ will have to do a memcpy()

Thirdly, if most of those modules will ever run in a threaded 2.0
environment they will have to learn better habits anyway. So someone
will propably update them sooner or later (and then we get much
cleaner binary pluginablility for 2.0 in any case :-).


Dw


Mime
View raw message