Re: URIs & Namespaces

hello john.

John Cowan wrote:
>> that's like the magicmail.org. if almost by definition you say not to 
>> access the http server and there are some fixed semantics to the uris in 
>> this domain, why bind all of that to something as fleeting as a 
>> registered domain name that belongs to somebody and, to understand this 
>> magic domain, has to be hardcoded or configured into any software 
>> processing these uris?
> On the contrary, I *do* say to access the HTTP server if you want to
> know what's going on.  If you merely want to understand names, however,
> you don't need to access it: you can recognize the meaning of the URI
> directly.

so by visiting http://magicmail.org/, how do i know what's going on? do 
they let me read the spec? in the end, if i want to send email, i must 
know smtp or have a program that knows how to do it, and then i use the 
email address that's appended to http://magicmail.org/ for composing an 
smtp message. nothing in http://magicmail.org/ itself lets me do that 
(apart from the fact they let me read the spec), and in fact, it is even 
harder to understand what's going on because it actually is a different 
communications mechanism, smtp, hiding behind some http facade. if 
magicmail.org is hijacked by some spam king and the good guys move on to 
magicmail.net, all my previous stuff breaks.

>> - magicmail.org or geo.example.com could set up a web page that informed 
>> people about the semantics that are built into the domain name. to be 
>> able to use the uris, you must then understand and implement the 
>> semantics, and associate them with the domain name.
> Exactly.
>> What that saves you, 
>> essentially, is looking that up in the iana uri scheme registry (where 
>> you would have to go if you encountered an unknown uri scheme).
> That would be fine if there were any such reliable registry.  There isn't.

really? you surely know http://www.iana.org/assignments/uri-schemes.html 
and there must be some reason why you don't think that's a reliable 
registry. enlighten us.

> Quite so.  But the added improvement to interop by using http URLs, plus
> the lower deployment costs (getting new schemes into browsers worldwide
> is no trivial affair), are a small but noticeable improvement.

i am still unconvinced that binding resource semantics to a domain name 
and tunnel that through http is a good thing. not if you think that the 
semantics should be explicit and universally usable.

cheers,

dret.

Received on Tuesday, 11 December 2007 22:09:55 UTC