Re: Length of labels (RE: [idn] IURL vs URL, IDNS name vs DNS name)

["Donald E. Eastlake 3rd" <dee3@torque.pothole.com>]

In other words, it would be just about as easy, and much safer, to
define a new UTF8 or whatever label type for DNS as it would be to
fundamentally change the "ASCII" label type to have a different length
limit.  (Yes, I'm quite aware that arbitrary binary values can be put
in "ASCII" DNS labels but applications screw up royally on nulls and
the like and the case folding can screw you pretty badly.)

If you are willing to contemplate software mods to everything that has
to touch idn's, a new label type to add to the existing "ASCII" and
recently defined bit string "binary" labels is probably the cleanest
thing.

Donald

From:  Harald Tveit Alvestrand <Harald@Alvestrand.no>
Message-Id:  <4.2.0.58.20000214235943.0275d510@dokka.maxware.no>
Date:  Tue, 15 Feb 2000 00:05:55 +0100
To:  Jonathan Rosenne <rosenne@qsm.co.il>, idn@ops.ietf.org
In-Reply-To:  <NDBBIFGHGLJILJGHOEELKEIKCEAA.rosenne@qsm.co.il>
References:  <1046892284.950234664@p2.jck.com>

>At 19:00 11.02.00 +0200, Jonathan Rosenne wrote:
>
>>The UCS could be encoded using A-Z 0-9. On average each non-ASCII character
>>would require 3 to 4 characters. All we need to change is to allow longer
>>names in DNS and to provide a viewer which decodes these creatures back to
>>UCS or the local code page.
>
>Unfortunately the limit of 63 octets per domain name component is enshrined 
>in a 6-bit length field in RFC 1035, section 4.1.4 (the 2 upper bits are 
>used as a flag to indicate pointers or "something-elses"), so "all we need 
>to change" to get this length increased is every single piece of DNS 
>software on the planet.
>
>Not that we might not come to that anyway...
>
>                          Harald
>
>--
>Harald Tveit Alvestrand, EDB Maxware, Norway
>Harald.Alvestrand@edb.maxware.no