Re: MS Windows packages and modularity

[Allin Cottrell <cottrell wfu edu>]

On Tue, 21 Oct 2008, Tor Lillqvist wrote:

> > > Was there a compelling reason for this reversion?
>
> Not really. As has already been said, if/when the GDI+ -based 
> pixbuf loaders would be used, then it would hopefully be 100% 
> clear that it makes sense to build them as built-in in the 
> gdk-pixbuf DLL. 

Yes, that seems to make sense.

> But now when I was forced to revert to the libgpej and libtiff 
> -based loaders...

Sorry (just out of ignorance): who or what forced you to revert to 
the libjpeg- and libtiff-based loaders?

> there are two more or less equivalent choices, both which have 
> their disadvatages: either 1) build separate png, jpeg and tiff 
> loaders, which means there are more files to distribute for 
> people who want to minimize the number of files in the GTK+ 
> runtime and still be able to load jpef or tiff files, or 2) 
> build built-in loaders which means the png, jpeg and tiff DLLs 
> are required even for a GTK+ runtime used in a situation where 
> it doesn't use any png, jpeg and tiff files.

I'm not speaking from any great base of expertise, but just IMO:

1) "...more files to distribute for people who want to minimize 
the number of files in the GTK+ runtime..."

Why would anyone want to minimize the _number of files_ in a GTK 
runtime package, as opposed to the size in bytes of the package?

2) I suspect that most GTK apps (other than apps such as GIMP 
which traffic in images) have no use for JPEG or TIFF 
functionality.  This is really quite specialized.

> Would the best solution be to build the png, jpeg and tiff
> libraries as static libraries, and link them statically into
> gdk-pixbuf?

This would avoid the need for the app developer to package 
third-party DLLs, yet it still seems a bit awkward: it includes in 
the byte-count of the GTK distro specialized functionality that, I 
suspect, few will need.  (I'd place PNG functionality in a 
different category since in many ways it's treated as "native" in 
GTK.)

-- 
Allin Cottrell
Department of Economics
Wake Forest University