Re: URIs for the standard output and input

>On Tue, Jan 15, 2008 at 07:26:06AM -0800, James M Snell wrote:
> > Well stated. std:in, std:out and std:err are perfect; using
> > http://whatever is ugly and buys us little
Noah Slater replied:
>Well, URI schemes must be registered with IANA. You should contact
>them to see what they have to say. I would be surprised if they didn't
>suggest that you use HTTP instead.
>
>Given that you only need an identifer, not a protocol hint, there is
>absolutely no reason to invent a new URI scheme for this.

The process by which IANA registers a new URI scheme is documented
in RFC 4395 (aka BCP 115).  There was considerable discussion
during the publication of the document about the right approach
to registering new URI schemes.   One of the serious concerns,
expressed both by IETF folks who use URIs to indicate protocol
processing and by folks who use URIs to mint identifiers, was
that  many folks just created new schemes without registration.
There were some serious conflicts (e.g. in incompatible "dns"
schemes and "mms" schemes).    Having non-interoperable
definitions of these schemes is a big problem; having a plethora
of schemes has its own issues, obviously, but using the
registration process as a filter was not apparently working to
slow the creation of new schemes.

The current rules reflect a desire to have folks think carefully about their need,
and if they do find such a need to register the scheme they create to fill it, at
least provisionally, so that the broader community is aware that they
have done so.

It's a trade-off; many things in life are.

I hope that pointer and background helps, but please understand
these are my comments as an individual and do not in any way
represent the IANA or IETF.
				regards,
					Ted Hardie

Received on Tuesday, 15 January 2008 17:28:24 UTC