On Sat, 2004-05-22 at 15:21, Jim Lyon wrote:
In previous messages, Andy made two useful points. (I'm not sure, but I
think that Doug may have been making one or both of them too.)
1. The right namespace for us to use is defined by BCP 81 (RFC 3688).
(If I read it correctly, our namespace and schema ought to be something
like "urn:ietf:params:xml:schema:marid-1".) IANA will register this
name, along with the complete contents of the schema.
2. We can employ various hacks to obviate the need to store the above
string in DNS records.
I agree with both of these.
In the spec that I'm writing, I'm currently planning to say that the
relevant XML document is formed by taking the contents of a DNS record,
then prefixing it with
<?XML VERSION="1.0" CHARSET="US-ASCII">
<root xmlns="urn:ietf:params:xml:schema:marid-1"
(possibly define other useful namespace prefixes here) >
and suffixing it with
</root>
A MARID token obviates a need for this entire header. The character
set should be part of the specification. The version number can be
understood to follow the token. Define this text in the draft and
simply indicate this text has been supplanted by the token. This
includes ALL URI information of any form. Otherwise, the camel's nose
is still under the tent.
The TXT record example for this entire worthless 97 byte header becomes
10 bytes:
<MARID-1 ...>
Only XML script defined within the standard SHALL be allowed as tokens.
Otherwise, when and where are these definitions created?
Otherwise, how does a parser handle a dynamic set of definitions without
also using HTTP?
Otherwise, how many templates are to be cached?
<snip>
-Doug