GSCUPS and GNUstep Printing

[Chad Hardin]

Now that GSCUPS is coming along fine I'm starting to think more of how 
it can be put into GNUstep.  I'm not sure that a printing backend is 
the way to go, it seems overly complicated.  Correct me if I'm wrong 
but basically the only things we have to worry about are:

Doing things the UNIX way:   Parsing PPDs, using lp(r) or cups.

And doing things the Windows way:  However the heck that is, I doubt it 
has anything to do with reading PPDs, I may be wrong though, wouldn't 
be the first time.

So maybe this is more of an #ifdef #elif type thing....  Comments?  How 
do people feel about making CUPS the sole supported printing system on 
UNIX machines?  If people really object to that, then there could be 
three setups:  UNIX like it is now, UNIX with CUPS, and Windows.

More on GSCUPS.  While it is true that GNUstep could work fine using 
CUPS without actually being aware of CUPS (because it provides the 
usual lpr type executables, etc), using the CUPS libraries has  two big 
advantages.  One is that CUPS can maintain and handle the PPDs, it can 
actually send the PPDs of the printers it serves to the client via 
HTTP.  That's one big headache gone.  You have to use libcups to get 
that.

The other thing is that CUPS has its own PPD parser and it turns the 
PPD file into a very fine-grained collection of c language structs.  Of 
course GSCUPS represents these as objective-c Objects.  So, we have a 
very good way to handle PPDs, for free.  Alternatively, the current PPD 
Parser could be used with the PPD file received from CUPS.  Obviously 
you need libcups for that too.

Right now GSCUPS is actually two things rolled into one.  It is a set 
of classes for doing simple querying of printers and the actual 
printing and a much more complicated set of classes for interacting 
with CUPS servers for the administration of the printers (which can 
also include printing, it's lower level stuff).  I think that if GSCUPS 
does get put into GNUstep, it should get split into two frameworks; the 
more complicated server administration part should be a separate 
framework.  If that is the case, then I'm talking about a very small 
patch for GNUstep itself.

By writing GSCUPS I've become pretty up to speed with CUPS, so I could 
probably add CUPS support for GNUSTEP using libcups directly, no need 
for an objective-c wrapper (If that would bother you all)

Comments?

Chad