W3C home > Mailing lists > Public > uri@w3.org > September 2013

Re: limit to length of scheme name?

From: Roy T. Fielding <fielding@gbiv.com>
Date: Tue, 3 Sep 2013 09:25:35 -0700
Cc: =JeffH <Jeff.Hodges@KingsMountain.com>, "uri@w3.org" <uri@w3.org>
Message-Id: <57F619F5-9FE1-463F-B89D-C06DA3F87A46@gbiv.com>
To: Gannon Dick <gannon_dick@yahoo.com>
It isn't an idea --- it is a fact.  There are no standard limits on
the length of a URI scheme for the same reason that there is no
standard limit on the length of a URI, and it makes no practical
difference whatsoever to any implementation worth talking about.

If you don't have control over the sender, length limitations on
received string values are irrelevant.  Protocols that define such
limitations are actually defining implementation capacity limits,
not protocol limits, and they invariably result in security holes
because inexperienced programmers believe in protocol fantasies.

In any case, the length of the name is irrelevant because a URI
processor is looking for the ":" delimiter and may have to scan
the entire reference before knowing whether it begins with a
scheme or a relative path.  Once a scheme is found, comparison
(for handling) is to existing schemes known to the recipient, so
there is no benefit to be obtained from very long schemes unless
the recipient is stupid enough to assume anything longer than N
characters is a path.

And, of course, any URI for pi would use its symbolic name as
the identifier (that's why we call it pi).

....Roy


On Sep 3, 2013, at 5:46 AM, Gannon Dick wrote:

> go look here for alternate view: http://www.loc.gov/marc/organizations/orgshome.html#struct
> http://tools.ietf.org/html/rfc4647 too.  The idea of unlimited scheme names does not sit well with me, personally: imagine a URI for the number pi.  PI = one number =one URI, pi is irrational with boundless digits.  Location (on a number line) and identification are different properties.
> --Gannon
> 
> From: Roy T. Fielding <fielding@gbiv.com>
> To: =JeffH <Jeff.Hodges@KingsMountain.com> 
> Cc: uri@w3.org 
> Sent: Tuesday, September 3, 2013 3:14 AM
> Subject: Re: limit to length of scheme name?
> 
> There is no limit on scheme name length.
> 
> ....Roy
> 
> On Sep 2, 2013, at 6:00 PM, =JeffH wrote:
> 
> > Hi,
> > 
> > I've perused RFCs 3986 and 4395, and do not see any length limitation for scheme names -- none is specified?  (not a criticism, just a yes/no(-go-look-here) sort of question)
> > 
> > thanks,
> > 
> > =JeffH
> > 
> 
> 
> 
> 
Received on Tuesday, 3 September 2013 16:25:59 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:16 UTC