Re: XML Property List code in GNUstep Base

[Richard Frith-Macdonald]

> On 14 Jul 2015, at 18:18, Stefan Bidigaray <address@hidden> wrote:
> 
> On Tue, Jul 14, 2015 at 12:58 PM, Richard Frith-Macdonald <address@hidden> 
> wrote:
> 
> > On 14 Jul 2015, at 17:04, David Chisnall <address@hidden> wrote:
> > I’d also be interested to see what NetBSD’s XML plist library does.  I 
> > believe that they just use the Apple DTD, which might be a better option 
> > for us - it’s easier for code consuming plists to validate that the DTD 
> > string is the same than to fetch and validate that the DTDs are equivalent.
> 
> Yes, the format (and presumably the dtd document location) has been stable 
> for several years now … I suppose there’s no reason we can’t point to it.
> 
> I would argue that providing our own DTD file is a more prudent choice.
> 
> For one, it provide a certain level of autonomy.  For example, is the 0.9 
> version of Apple's DTD still available?  What guarantees that version 1.0 of 
> the file will be available 5 years from now if the version is bumped?

I think that was the rationale for going our own way originally (especailly 
since the 0.9 designation seemed to imply that change was imminant) but it 
seems we’ve had the dtd document missing from the gnustep website for a long 
time before anyone even noticed, so the lack of a dtd does not seem to be a 
problem in practice, and it probably doesn’t matter if apple take it off their 
website.
 
> A XML plist parser does not need to be a full XML parser, and can skip the 
> DTD validation altogether (which is what I plan on doing in CoreBase).  The 
> format is brain-dead simple and a XML plist parser can go through data 
> efficiently without ever having to validate, returning NULL on a syntax error.

Yes, that seems to be another argument for not bothering to keep our own 
version of the DTD: people aren’t going to check it anyway.