[Mapserver-dev] FastCGI Support in MapServer

[Sean Gillies]

On Sep 17, 2004, at 8:33 AM, Frank Warmerdam wrote:

> Sean Gillies wrote:
>> Frank,
>> I'd like to help you with the standard stream wrappers.  This could 
>> help
>> us share the WMS dispatching with mapscript
>>   http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=670
>> Would mapservObj be changed to store input and output filehandles, or 
>> would
>> they be stored in mapObj?
>
> Sean,
>
> In my initial thinking there would be global (essentially hidden)
> file handles.  They wouldn't be in the mapservObj or map.  What I 
> *don't*
> want to do is have to start passing extra objects into low level 
> places that
> never needed anything other than stdio functions before.  So, my theory
> was to use the normal stdin, stdout and stderr FILE*'s as "identifiers"
> that would be used in the FastCGI jacket functions to so select the 
> appropriate
> FastCGI file handles.
>
> Reading over bug 670 it isn't entirely clear to me what sorts of IO 
> you are
> needing to virtualize.  Are you wanting to be able to capture "stdout" 
> output
> of stuff like the getcapabilities generator within a script to 
> manipulate?
>
> We could indeed look at a more ambitious approach to 
> stdin/stdout/stderr
> virtualization with some sort of file objects passed around to control
> redirection in a manner that could be map, thread or mapservObj 
> specific,
> but this would make what I need to do substantially harder.
>

Frank,

Yes, I want to be able to get WMS capabilities as a string for starters.
In bug 670 I was asking Daniel if we could replace printf everywhere 
with
fprintf ... or create another layer that works on explicitly defined
filehandles instead of stdin/stderr/stdout.

I want to avoid globals with mapscript, so maybe our needs are 
divergent.

Sean