Re: [syndication] RSS 0.91 DTD: So what now?

[Mark Nottingham <mnot@mnot.net>]

Howdy from Hong Kong ;)

I'd be the first person to say that distributed systems, redundancy, etc. is 
cool, and encourage those kinds of solutions. However, I don't know if they're 
really appropriate here; just because a URL is used to identify a document, it 
doesn't mean that it needs to be fetched (located) every time it's referenced, 
especially if it's something as critical and unchanging as a DTD. 

I would imagine that a Good validating parser would allow local configuration 
to determine where to look for the DTD, rather than automatically going to the 
network. In effect, a local, manually-populated cache. This is much more robust 
in several ways than going to the network, locating the resource every time you 
use it, and hoping that the authority that publishes it doesn't change its mind 
while you still need it (as has happened here).

If it really needs to be locatable on the network, there's some really 
interesting work in the IETF regarding URNs happening right now; this would be 
an ideal opportunity to play with them.


Quoting Dave Winer <dave@userland.com>:

> There's another possibility, we could get a set of replicated
> directories
> going, one lightweight centralized and replicatable node, that keeps
> lots of
> replicated servers updated.
> 
> Of course this is just another way to look at the xmlStorageSystem spec
> we
> published a few weeks ago.
> 
> http://www.soapware.org/xmlStorageSystem
> 
> Ever the evangelist..
> 
> Dave
> 
> 
> 
> ----- Original Message -----
> From: "Rael Dornfest" <rael@oreilly.com>
> To: <syndication@yahoogroups.com>
> Sent: Friday, April 27, 2001 3:27 PM
> Subject: RE: [syndication] RSS 0.91 DTD: So what now?
> 
> 
> > Hi,
> >
> > : Rael, I think the first thing to do is to check and double-check the
> DTD
> > : that's there and be sure it's the right one. It seems that should
> be
> > : relatively easy to determine. I would hate to think that I got the
> wrong
> > : copy or there's a transmission error, and somehow that gets
> replicated.
> >
> > It does indeed seem to be the right version; at least it jives with
> the
> DTD
> > in XML::RSS's package.
> >
> > : Second, I think the idea of a directory of replicates is a great
> > : one. I also
> > : think that now that we have the chance, we should include the url
> to
> that
> > : directory in a comment in the DTD, in case we go belly-up, or get
> > : taken over
> > : by AOL, or whatever.
> >
> > So how do people reference this DTD in their feeds?  What if that
> server
> is
> > down for the few seconds a parser is working with a feed?
> >
> > There's going to be a whole lot of shuffling as people are changing
> their
> > RSS files to use the DTD that's most available at any one point.
> >
> > : Beyond that, I don't have any more ideas.
> >
> > RSS needs to, 1.0 or otherwise, move away from the static DTD model
> and
> > reliance on a single point of availability (read: inevitable
> failure).
> >
> > Rael
> >
> >
> >
> >
> > Your use of Yahoo! Groups is subject to
> http://docs.yahoo.com/info/terms/
> >
> >
> 
> 
>  
> 
> Your use of Yahoo! Groups is subject to
> http://docs.yahoo.com/info/terms/ 
> 
> 
>