>>>>> On Wed, 30 Mar 2005 09:18:15 -0800,
>>>>> Bill Fenner <[email protected]> said:
>> 3. the parser passes "fe80::1_de0" to getaddrinfo(), and gets a
>> sockaddr_in6 structure (whose sin6_addr member is "fe80::1" and
>> sin6_scope_id member is the link ID corresponding to interface
>> "de0"). The browser uses the sockaddr_in6 structure with
>> connect(2) to connect to the remote web server.
> I was assuming that the parser would have to turn fe80::1_de0 into
> fe80::1%de0 before passing to getaddrinfo(), so would have to parse
> out the zone explicitly.
Ah, okay, thanks for clarifying that.
But...
> RFC 4007 says that in the general case, applications shouldn't have
> to know about hacking zones off - but this is format is explicitly
> for a special case.
I'm feeling I'm confused again...please let me go back to question 0
"Should we solve this problem at all?". I said answering YES to
this question was against the general sense of RFC4007. Regarding
this point, the above logic seems to say the delimiter conflict issue
makes it (being against the sense of RFC4007) a non-problem, but it
does not seem very logical to me.
The essential point is, at least to me, is that we did not want to
force applications (like URI/URL parsers) to be aware of scope zones
and/or the dedicated syntax for scoped addresses. This point doesn't
change regardless of how much the application needs to work to deal
with scopes (i.e., whether it only needs to split the address part or
it also first needs to convert _ to %). In fact, if we can now allow
this as a special case, I cannot understand why people were so
enthusiastic about killing site-local addresses -in my understanding,
one major reason was they did not want to force applications to know
details about scopes and/or zones for behaving correctly.
What we should provide regarding question 0 is, IMO, a convincing
reason why we dare to be against the general sense of RFC4007 without
relying on the fact that converting _ to % would make the application
more complex.
BTW, if we can really agree on being against the sense of RFC4007 for
the particular usage of URI/URL, option 3 in
draft-fenner-literal-zone-01 (just use "%") rather seems to be the
most appropriate resolution. Aside from the point of "syntax police",
the difficult issue of this option in practice is that we cannot be
sure if we can enforce the special case (against the URI/URL syntax)
to URI/URL parsers (or implementors). But since we'd now be going to
enforce a special exceptional rule (i.e., enforcing to extract the
"pure address" from the "address+scope zone"), wouldn't it also be
feasible to force the parser to regard '%' in "[...]" as an exception
to RFC3987?
(Note: my primary position is still I'm not convinced of the answer to
question 0. The last paragraph is only meaningful if we can agree
that the answer is YES.)
JINMEI, Tatuya
Communication Platform Lab.
Corporate R&D Center, Toshiba Corp.
[EMAIL PROTECTED]
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
