RE: DNS Draft

Alas Roy there is what there is and there is what there should be.
Currently all there is, is the ability to register the schema, there is
not a way to register down level paths. While I agree that registering
paths would have been the best way to handle this situation, matters did
not turn out that way.

Either way, however, we do need a way to keep private schemes from
colliding and the draft in question appears to provide one satisfactory
to all players.

			Yaron

> -----Original Message-----
> From:	Roy T. Fielding [SMTP:[email protected]]
> Sent:	Thursday, September 04, 1997 12:30 AM
> To:	Yaron Goland
> Cc:	'[email protected]'
> Subject:	Re: DNS Draft 
> 
> >The reason I didn't do it that way is that neither Netscape nor
> >Microsoft can support it properly. Both systems do registration based
> on
> >the protocol name, not on the path. As such we would have a perfectly
> >reasonable standard that doesn't solve the problem posed by already
> >deployed code.
> 
> I guess I may be thinking of a different implementation model than
> is actually in use.  The "sensible" way of handling a new scheme is
> to introduce a new handler for the scheme within a registration table.
> So, you register a handler for the generic "path" scheme which looks
> at the URL, checks for "path:/com/microsoft/" and says not implemented
> if no match, otherwise it drops that part of the identifier and uses
> the rest to do whatever application-specific thing you wanted to do.
> 
> I don't see how that is any different for deployed software than
> defining
> separate handlers for each scheme.  Is it impossible for one handler
> to call another, or is there some other constraint that I just don't
> know about?  I'd look at the source, but ...
> 
> OTOH, I have never been concerned about protecting the scheme
> namespace
> from collisions.  They tend to be temporary phenomena.
> 
> ....Roy