Re: [upnp-sdk-dev] release 1.3.1, rpm, ixmlPrintDocument

[<tur...@ea...>]

John,

> First of all Remi, thank you for your work and releasing version 1.=
3.1.
> I've downloaded it and did some testing. I also reviewed the work y=
ou
> did to convert building with automake, libtool, pkgconfig etc. It l=
ooks
> excellent and is a huge improvement!
>  =20
thanks !

> On the issue of installation, I see that the libraries for threadut=
il
> and ixml are installed in the system in parallel with upnp. I wonde=
r if
> in the long run that's a good idea. (...)
>  At some point it might be
> worthwhile to link these libraries into libupnp and hide their symb=
ols.
>  =20
except that ixml has its own documented interface (see the libupnp-do=
c=20
package), and normally is used directly by programs interfacing with=
=20
libupnp.I think it would not be practical to hide it. May be merge it=
,=20
but that would break programs linking with the 3 libraries.

>
> In the UPnP SDK ixmlPrintDocument had been #define'd to be ixmlPrin=
tNode
> which prints XML elements. The fix I believe is to have to independ=
ent
> functions, ixmlPrintDocument and ixmlPrintNode (same for the toStri=
ng
> versions) where the Document functions include the prolog and the N=
ode
> functions omit it. The attached patch (which I'll also post on the =
SF
> site) creates the new Document functions which prepends the prolog =
and
> replaces every instance of ixmlPrintDocument with ixmlPrintNode bec=
ause
> the two are now no longer interchangeable (except in configure_urlb=
ase()
> in the file urlconfig.c, because here you do want a document).
>  =20
you are right, that's the way ixmlPrintDocument should have been=20
implemented from the beginning. However, I have seen programs using t=
his=20
existing interface as it is, and manually adding the prolog.
This fix would silently break this interface. To keep backward=20
compatibility : wouldn't it be preferable to add a new function (e.g.=
=20
ixmlPrintDocumentWithProlog or something like that), and document the=
=20
deficiency in the old function?

Also : another user suggests to implement a check in the libupnp, in=
=20
order to add the prolog in the device description document if it is=
=20
missing :
   =20
http://sourceforge.net/tracker/index.php?func=3Ddetail&aid=3D931724&g=
roup_id=3D7189&atid=3D107189
I don't know if this other patch is correct or not, but we could=20
implement this idea too, in addition the the ixml fix.

What's your view ?

R=E9mi