Lists Home |
Date Index |
----- Original Message -----
From: "Tim Bray" <firstname.lastname@example.org>
> Joshua Allen wrote:
> > On the other hand, it is exceedingly poor form to use identifiers from
> > the HTTP scheme for things which you don't intend to be dereferenced
> > (via HTTP synchronous GET, no less).
> You might be on to something here.
> How about for things that you don't have any representations for right
> now but plan to in the near future? How about things that you don't
> have any way of representing right now, but you might someday? What are
> some things that fall into the category "which you don't intend to be
> dereferenced"? -Tim
There is nothing that says a resource cannot have more than one URI. For
instance, suppose I have a new idea. To distinguish it from all other ideas
I have, I assign the following URI to it:
The above URI only identifies a representation of the idea. At this point,
I have such a rough idea that I can't even effectively explain it to anyone
else. Later, I have enough of the idea to be able to type it into a file
and assign the following URI to it:
This URI now points to a different representation of the same resource. I
could have used the second URI all along, but it would have been able to be
dereferenced. When I had something that could be dereferenced via http, I
assigned a new URI using the "http" scheme to allow the dereferencing.
The thing that struck me was what Joshua Allen alluded to his next post
(well, I think he was alluding to it) and what the above example
If we all start using strings that match URI's using the "http" scheme that
are not actually meant to be dereferenced/resolved/whatever, then we start
muddying the "http" scheme global namespace, where the "common" practice is
to be able to derereference URIs (ignoring server problems, etc.). If you
muddy it enough, people stop paying attention to it. After all, if you
attempted to go to the following URIs:
and none of them dereferenced (and they don't), how likely are you to
dereferenced (and it does)?
Sure, you could say that I am being extreme. You could say that it's very
doubtful that there will ever be so many non-dereferencable http-scheme URIs
that this would ever be a problem. And all I know is that, with as few of
these that *are* already out there, they are causing never-ending threads
like "URIs harmful".