- From: Paul Prescod <paul@prescod.net>
- Date: Tue, 02 Apr 2002 14:22:37 -0800
- To: Tim Berners-Lee <timbl@w3.org>, www-tag@w3.org
Summary: We agree that URIs should not specify behaviour, even "default" behaviour like "send message" or "mount NFS file system". What we disagree on, it seems, is whether there should be URIs for things that are not in the "information space" (i.e. GET-able in some sense). I think no, we can reference them without giving them URIs. Tim Berners-Lee wrote: > >... > > The mailto: schema name was badly chosen, but the concept is > sound as originally defined. It was intended simply to be a space > in which to put all the internet email addresses, which are called > mailboxes. A mailbox is an abstract thing, related to email messages > by (for example) To: From: and Cc: feilds but also used in many > other situations. It also normally has a relationship with the social > entity > -- typically a person or group --which owns it. > > The unfortunate choice of name "mailto", my fault, has lead browser > designers to think that it specifies an action. When you assign a part of the namespace to an abstract thing you kind of invite abuse. Browser manufacturers are *required* to map URIs to actions. When the end-user clicks they need to do something. If you don't tell the browser vendors what action to invoke then the first popular action will become the defacto standard. >... > The disadvantage is that you have just lost your address space. You have > lost > the ability to simply and unabiguously refer to an email address. > Instead, you have to dererence lots of tiny documents just to compare two > email addresses for equality. (You may be looking at email offline!) I disagree. Just as with any other URIs, you can either use pointer or reference equality. The mapping of my email address into the web space might be: http://www.prescod.net/myemail.mbox There is no reason for anybody else to define another URI for my email address (although they could), so pointer equality will suffice for most cases. Fine, but that doesn't imply they need to be part of the URI space. IMO, the URI space is for addressing the information space! Things that use protocols that do not support GET can be referenced through non-URI mechanisms. That's already what RealAudio does (for example). They need to do that because of the difficulty of deploying new URI types, but it is also a good idea because it makes it easy to associate metadata with the resource. Paul Prescod
Received on Tuesday, 2 April 2002 17:26:32 UTC