Re: Getting full URI to the server

Thus wrote: "Roy T. Fielding"
>   Host: foo.bar.org
>
>since that is the only thing interesting which was left off the request.
>[Before anyone mentions it: no, the port number is not interesting].

No, but the fragment identifier (i.e. the #foo that may go on the end
of a URL) could be there, and could be interesting in some cases.
There had been some discussion of extending fragments to allow
searches for particular strings, line references, that sort of thing;
including it in the request line could allow servers to do
preprocessing or only send part of the document if such is
appropriate, which could in theory be more efficient.  I don't find
this compelling, but it is another potential advantage of sending a
full URL.

>Note that it has no other advantage, as the only non-distinguishable link
>to that server is the root.

I wouldn't quite agree there.  Other site-level conventions
(e.g. /robots.txt, /site.idx, a file for site-level description
information if one is made, etc.) could also benefit from this header.
It also could permit a smoother migration of name; for example, if you
need to change the name of your server, this header could permit the
old and new names to be mapped together briefly but permit logging of
which Referer: URLs are still pointing to the old name.

That said, I'm inclined to agree that the benefits are not terribly
overwhelming.

- Marc
--
Marc VanHeyningen  <URL:http://www.cs.indiana.edu/hyplan/mvanheyn.html>

Received on Sunday, 12 February 1995 06:30:31 UTC