Re: now://example.org/car (was lack of consensus on httpRange-14)

> That's true, of course.  But I think it misses the point that Micah and
> others are tying to make:  so much of the web folklore, and the client
> programs, and the practical experience of people who have no idea what
> "HTTP" in the URL means, leads all but uber-geeks inexorably to the
> conclusion that "HTTP means I click on it and get something back."

In my 8 years of experience as co-author of the HTTP and URI
standards and core developer for the Apache httpd Web server, I have
never once met a user of Web software who was confused by the use of
an http URI as an identifier.  Users know that a link defined by an
anchor does something useful when you "click on it", a link defined
by an img src embeds an image in a page, and an address typed into
a Location bar has the same effect as an anchor.  They never see how
http URI are used in other respects and simply do not care, nor do
they have any expectation that "http" implies network request, let
alone an HTTP request.  It is simply an identifier that the browser
will use according to the context of its use, not according to some
implied nature of its syntax prefix.

There is no point in arguing this further.  I have to explain what
it is that I do for a living on a regular basis and from experience
I can tell you that non-technical users don't know what the "http"
stands for in a URI.  They don't even know that HTTP exists.  Nor
should they, since the Web architecture allows any URI scheme to
be used in any of those places, thus enforcing a disassociation
between the scheme and any action that might be implied by that scheme.

It is actively harmful to the Web architecture to suggest that the
URI scheme -- any URI scheme -- implies an action.  It must not.
It never has and never will because the whole generic interface
depends on the separation of concerns between method and identifier.

....Roy

Received on Wednesday, 9 October 2002 00:07:33 UTC