Re: IRI length limits (ticket #37 => #46, #47, #48)

Hello Ted,

Many thanks for your comments.

On 2010/09/29 3:18, Ted Hardie wrote:
> Hi Martin,
>
> Some comments below.
>
> On Tue, Sep 28, 2010 at 3:08 AM, "Martin J. Dürst"
> <duerst@it.aoyama.ac.jp>  wrote:

>> #46: Normative length limits
>> (http://trac.tools.ietf.org/wg/iri/trac/ticket/46)
>>    I personally propose that we close this issue without action, because
>>    I was unable to find any such limits in RFC 3986 (URIs), and do not
>>    think such limits make any sense.
>>
>
> It's not clear to me whether you are closing the normative  issue for
> domain names or for IRIs in general.

Just to make sure, I'm not closing it, I'm just proposing to close it.

> RFC 1034 has some normative limits;
> e.g. a label is between 0 and 63 octets in length.  I agree that we should
> not introduce new limits, but reference to the ones which do exist still
> seem to me necessary.

I agree in principle. But the 63 octet length limit for domain names 
does not directly affect IRIs, and therefore cannot be a normative 
limit, only an informative one. You give an example of that further 
below, but to quote from the URI spec (RFC 3986, see
http://tools.ietf.org/html/rfc3986#section-3.2.2):

    This specification does not mandate a particular registered name
    lookup technology and therefore does not restrict the syntax of reg-
    name beyond what is necessary for interoperability.  Instead, it
    delegates the issue of registered name syntax conformance to the
    operating system of each application performing URI resolution, and
    that operating system decides what it will allow for the purpose of
    host identification.  A URI resolution implementation might use DNS,
    host tables, yellow pages, NetInfo, WINS, or any other system for
    lookup of registered names.  However, a globally scoped naming
    system, such as DNS fully qualified domain names, is necessary for
    URIs intended to have global scope.  URI producers should use names
    that conform to the DNS syntax, even when use of DNS is not
    immediately apparent, and should limit these names to no more than
    255 characters in length.

While this gives the number 255, this seems to be only advisory, not 
normative.


>> #48: Can schemes set specific length limits? Should RFC4395bis say something
>> about this?
>>    My current thinking is that obviously, schemes can set some length
>>    limits, because they can define syntax restrictions. Also, they
>>    could do that in silly, counterproductive ways. But this all seems
>>    to be rather obvious, so I personally think that we may not need
>>    to say anything on this topic.
>>
>
> I think it should be noted, probably in RFC 4395bis.  Some schemes
> do not use the domain name system for the authority section, to take
> one example, but might have similar restrictions on their equivalent
> to "label length".    If you look at the draft I put forward a while
> back p2p URI pointers (see
> http://tools.ietf.org/html/draft-hardie-p2psip-p2p-pointers-01
> now expired), you'll see an overlay-pointer-uri.  That was generalized
> sufficiently that there was no length limit, but I can easily imagine specific
> schemes limited to the maximum production of the mechanism for producing
> the node-id (commonly 128 bits).
>
> It may be obvious to us, but I'd rather err on the side of noting the
> possibility.

Fair enough.

Regards,   Martin.

-- 
#-# Martin J. Dürst, Professor, Aoyama Gakuin University
#-# http://www.sw.it.aoyama.ac.jp   mailto:duerst@it.aoyama.ac.jp

Received on Wednesday, 29 September 2010 10:14:58 UTC