Re: a bad idea (fwd)

Marcus E. Hennecke <marcush@crc.ricoh.com> wrote:
|It would break Netscape. It simply takes everything in the quotes to be
|the URL, including spaces. If you try:
|<a href="http://foo.com/ http://bar.com/">Follow this link</a>

It didn't used to break Navigator, at least when I tested it last year.

If what you say is true, that means that Navigator's interpretation of a URI is
noncompliant per RFC 1738, Section 2.2, and this convention would expose an
already broken implementation.

|The problem with many of the proposals offered here is that they are not
|general enough to allow the browser proper selection of the most desired
|URL. Different URLs may differ in access time, proximity, language, data
|format (e.g. GIF, JPEG, PNG), compression (e.g. none, zip, gzip, etc.),
|size, charset, etc.

Part of this could be solved in HTTP with multiple attempts at content
negotiation for each alternative and need not be expressed in the container
document.

You outline a very large problem that, in order to be addressed most fully,
would need a solution along the lines of the URN/URC efforts.  Attempting to
overload HTML with the descriptive features of reasonably complex metadata
scheme is a Bad Idea.

One can't assume that the maintainer of an HTML container would be in control
of all objects and their replicates referenced objects can change, and if their
mutable characteristics become part of the container doc, then caching can be
easily defeated.  Further, the task of keeping a container doc in sync with the
status of multiple disparate objects is a maintanance nightmare that I'd rather
avoid.

If resource replication going to be addressed in HTML or its successors, it
should be kept brain-numbingly simple.  Otherwise it should be fully fleshed
out in another forum and available to the user agent through a LINK.

-marc

--

Received on Monday, 15 July 1996 16:28:45 UTC