Re: [EAI] The mailing list draft

[John C Klensin <[e-mail]>]

Martin,

A few comments (personal, not a co-chair except as noted).  I'll
let John L respond to the others and get a new draft posted when
he is ready.  

<co-chair hat=on>
WG members: some of this is controversial.  If you have
comments, make them.  WG LC on this document will start as soon
as John gets a new version posted (even if only to avoid WG LC
on an expired draft).
</co-chair>

--On Tuesday, June 19, 2012 16:08 +0900 "\"Martin J. Dürst\""
<[e-mail]> wrote:


> 2.2 in general: It occurred to be that one way of implementing
> this is to have two mailing list addresses, or in some way
> almost two mailing list, one old-style and one EAI, where
> submitted messages are automatically shared. Or is this a bad
> idea for some reason? If not, a more explicit description
> might help.

I'm not certain I understand this.  If I do, it is either
equivalent to advise to simply not permit UTF-8 addresses in
mailing lists or any EAI-requiring messages to be posted to
them) or having the expectation of having lots of messages
delivered twice.  Another reading would take us back to
requiring that every UTF-8 address be accompanied by an
alternative, 5821-conforming, address, which takes us right back
to all of the issues associated with dual addresses and
in-transit downgrading, including the security-threat issues,
that we agreed to drop.  Or am I misunderstanding the suggestion?

> 3.1:
>     "The current specification for
>     mailto does not permit unencoded UTF-8 characters":
> That's wrong. RFC 6068 (http://tools.ietf.org/html/rfc6068)
> allows 'raw' Unicode characters everywhere except the LHS of
> the address,
> and even for that part, RFC 6068 already "does the right
> thing",
> although with a "reserved" caveat. Of course, 'raw' Unicode
> can only
> be used in the IRI form, for URIs, %-encoding based on UTF-8
> is needed.

Sadly, "%" has a different historical interpretation for email
local-parts than it does in URUs.  That isn't a problem as long
as all of the decoding and encoding rules are followed
carefully.  But, having just had another bad experience that
appears to be due to moving from email address to MAILTO URI to
email address fouling up "+" in a local part, I have no
confidence at all about proper handling of "%".

You and John note another aspect of this issue in the text and
elsewhere in your message.

>...
>    "Note that discussion  on
>     whether internationalized domain names should be
> percent-encoded or
>     puny-coded, is ongoing; see [I-D.duerst-iri-bis]."
> This point is no longer under discussion in the IRI WG.
> There's no essential change from RFC 3986, where these two
> forms already are allowed.

<co-chair hat=on>
I strongly prefer that we not drag the IRI debate into this WG.
"No longer under discussion in the WG" is different from
"approved in the IETF.  If we do have to make an IRI discussion,
or anything that is contingent on what that WG is doing,
prerequisite to making progress on the EAI mailinglist document,
I will propose to abandon that document until and unless the IRI
work produces IETF consensus Standards Track documents.
</co-chair>

>...

Thanks for the careful reading.

best,
   john