Re: apache (fwd)

[Brian Behlendorf <br...@organic.com>]

On Mon, 30 Sep 1996, Dirk.vanGulik wrote:
> *IF* we are going to make some additional stubs; and be backward compatible; these
> are the ones I'd like to add
> 
> 	- Post config-read; so we can check consistency between
> 	  directives and abort the startup
> 	- Per child init
> 	- Per child die
> 	- Pre-handler; which asks round what could *potentially* be served on a URI; 
>           gets back a list of possible internal redirects with the mime-type/lang/charset
> 	  and quality values.
> 	- Up to 4 placeholders for future use ?	

Okay, this sounds fine.

I think we should do this *once*, and do it right, and my personal opinion is
that this is thus a 2.0 issue.  If anyone thinks it should be in 1.2, speak up
now.

	Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com  www.apache.org  hyperreal.com  http://www.organic.com/JOBS

Re: apache (fwd)

[Dean Gaudet <dg...@hotwired.com>]

If the first field of the module structure was a byte length (or count
of the function vectors) then it's easy to be backwards compatible
with old modules without pre-allocating slots.  The modules code simply
allocates a dynamic structure of the correct size and copies in the
static structure for any "old" modules with the wrong count/size.

Dean

In article <ho...@fully.organic.com>,
Brian Behlendorf  <ne...@hyperreal.com> wrote:
>On Mon, 30 Sep 1996, Dirk.vanGulik wrote:
>> *IF* we are going to make some additional stubs; and be backward compatible; these
>> are the ones I'd like to add
>> 
>> 	- Post config-read; so we can check consistency between
>> 	  directives and abort the startup
>> 	- Per child init
>> 	- Per child die
>> 	- Pre-handler; which asks round what could *potentially* be served on a URI; 
>>           gets back a list of possible internal redirects with the mime-type/lang/charset
>> 	  and quality values.
>> 	- Up to 4 placeholders for future use ?	
>
>Okay, this sounds fine.
>
>I think we should do this *once*, and do it right, and my personal opinion is
>that this is thus a 2.0 issue.  If anyone thinks it should be in 1.2, speak up
>now.
>
>	Brian
>
>--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
>brian@organic.com  www.apache.org  hyperreal.com  http://www.organic.com/JOBS
>

Re: apache (fwd)

[ra...@vex.net]

On Mon, 30 Sep 1996, Dirk.vanGulik wrote:
>      - Per child init
>      - Per child die

I would love to have these two.  However, within the framework of the
API, would it be possible to also define some sort of data structure
which would let me, for example, open up socket connection to an RDBMS
and then have the file descriptor available to modules?

One of the big problems I have right now with building Apache modules
which connect to an RDBMS is that the startup time needed to establish
a connection to the backend RDBMS is much too long.  It is the same
startup procedure each time with 'nobody' logging into the RDBMS.

If we could have a pool of such persistant identifiers along with some sort
of registry check functions which would keep track of these identifiers, then
my life would be much simpler.

-Rasmus