[Prev][Next][Index][Thread]

Re: The furore over PUBLIC



Jon Bosak wrote:
> 
> [Len Bullard:]
> 
> | What about the case where the writer only wants to identify the type
> | and not the instance?  They must point to the registry/process where
> | the type is resolved to an instance.
> 
> My intuition is telling me that there is some helpful distinction to
> be made here between an FPI that's needed to identify which DTD a
> document is supposed to conform to and a generic public identifier
> whose purpose is to fetch external entities, but I haven't been
> successful yet in getting a grip on it.

Yes, and perhaps that is at the root of this.  We may be wanting 
to *do* something different.  I need a way to identify links 
independent of the system location, but that is just indirection 
and perhaps not what the public ID achieves.
 
> Surely what we're thinking of when we specify a version of the DocBook
> DTD is only in the last resort an actual trip to the Davenport site to
> get a copy, but that always has to be the fallthrough or the citation
> is ultimately meaningless.

Yes.  It is my point that we may be confusing the *name* by which 
something is registered for the *process* and the process is what 
is wanted.
 
> | I agree that a public id looks like a comment until one notes it is a
> | formal comment with a legal requirement to be a registered type.
> 
> No, there's no legal requirement inherent in "-//Davenport//DTD
> DocBook V2.4//EN".  And no formal registration with an authority,
> either.  I think that you're on to something, but that's not it.

A legal requirement is what a contract can cite and a process 
or agency can enforce.  That is what I mean by that.  If two 
parties agree that the string you cite is a formal name for 
authoritative registration, it is the contractual agreement 
that makes it so.  Otherwise, a string is a string is a string.
 
> | Using it without such registration makes it a less powerful formality.
> | Such formal registration also has domains; the registry server is the
> | domain server.  A lookup table is a lookup table is a lookup table, so
> | I don't understand the problem here.  It looks like another RFC for a
> | resolution service and if so, then those that want it have to
> | implement it to use it.  Duh!
> 
> The people who have to implement it are the browser vendors.  They're
> not buying a pig in a poke, and I can't say that I blame them.

Nor would I, but that is their decision to make.  Others are trying 
to write requirements and they are customers who do have a say in 
what is wanted.  The vendors can choose to sell that to them or not, 
and they can choose to find a vendor who will.  That IS how the 
market works.  Again, Don't Fear The Reaper, even if it is currently 
dominant in the reaping business.

len


References: