Re: [upnp-sdk-dev] Fork of project UPnP

[John D. <jd...@re...>]

On Wed, 2006-05-17 at 12:21 +0200, qx...@gm... wrote:
> One main focus beside of the further development of it are porting the
> library to other operating systems so that it is able to become a standard
> for UPnP applications.

One thing I was never comfortable with was how the SDK implemented it's
own version of what is now otherwise common interfaces, for example XML
parsing and generation, but there are other examples too (e.g.
threading). There would be less bloat if one would leverage off of the
work of other projects by linking against existing libraries rather than
trying to do it all. For example let the XML folks worry about producing
the smallest fastest most robust XML code and let the UPNP folks become
familar with just one API that will serve them on other projects as
well.

I think I understand why the original developers choose to implement
everything themselves, at the time of the original development standard
libraries did not exist for many of these technologies, but they do now.
Perhaps one thing which could be done is to trim down and simplify the
project by utilizing other libraries.

I can think of one reason to keep the existing "do it all" approach, it
may provide the smallest footprint if embedded devices are considered a
primary target, but this would need a bit of investigation.

The other thing I might suggest, and perhaps I could help with, is that
I found the example code to be verbose and awkward. For my project I
started with the example code but it wasn't long before I had rewritten
much of it. This is of course all a matter of personal style and taste.
But for me I found it difficult to get at the essence of how one needed
to code using the library because I thought the example code obscured
things. I wonder if others had the same reaction.
-- 
John Dennis <jd...@re...>
Red Hat Inc.