- From: Bruce Lilly <blilly@erols.com>
- Date: Mon, 20 Jun 2005 15:25:36 -0400
- To: uri@w3.org
- Cc: Paul Hoffman <phoffman@imc.org>, Larry Masinter <LMM@acm.org>
On Mon June 20 2005 13:43, Paul Hoffman wrote: > At 10:24 AM -0700 6/20/05, Larry Masinter wrote: > >(URIC-encoded) 'addr-spec' is reliable > >(URIC-encoded) 'mailbox-list' MAY be used, but is often > > badly implemented > > This would lead to massive lack of interoperability of software that > creates or resolves mailto: URIs. Assume that Thunderbird implements > reading the new form, so Thunderbird users start to create mailtos > for themselves using it on their web pages. A non-Thunderbird-user > double-clicks on the link, their MUA launches, and malformed crap > appears in the To: field of the new message. The same could have been said when the change was introduced between RFCs 1738 and 2368. Note that in fact RFC 2368 (using RFC 822 EBNF) says: mailtoURL = "mailto:" [ to ] [ headers ] to = #mailbox [...] "#mailbox" is as specified in RFC 822 [RFC822]. and further noting that RFC 822 '#' ENBF syntax in fact specifies a list, updating to RFC 2822 and ABNF actually *reduces* the potential for interoperability problems by eliminating generation of constructs permitted by RFC 822 -- including, but not limited to, empty list elements permitted by '#' syntax and the liberal inclusion of whitespace, comments, and line folding explicitly permitted by RFC 822 (note that 2822 deprecates comments in address fields). The draft proposal as currently written, and as likely to be when revised according to comments will probably require resetting status to Proposed Standard. Documenting implementation issues is not required until elevation to Draft Standard, so worrying unduly about such issues now is probably premature (however, note that the comments on the draft specifically pertain to interoperability, including backward compatibility with RFC 2368 as well as clarifying interoperability-unfriendly ambiguities in the 2368 specification). That said, there's nothing wrong with an implementation note observing the fact that some deployed implementations may fail to handle the full 2368 or draft-under-discussion syntax. > I propose that breaking current MUAs is not a good idea in order to > allow comments and lists of names in URIs. Both were already allowed under RFC 822 syntax rules as used by RFC 2368! RFC 1738 used RFC 822 syntax for addr-spec, which also explicitly permits liberal use of comments, whitespace, and line folding. > Even creating a new scheme > would probably cause confusion, particularly if you name it something > like 'mailto-new'. I have heard no demand for this new feature, even > though it has been obvious for over a decade. Paul, I'm puzzled by that statement; as a 2368 co-author, could you comment on why the feature was introduced in RFC 2368? > >What do you mean by 'insecurely'? > > It would probably be easy to create "new-format" mailto: strings that > would cause current MUAs to crash in interesting fashions. It is useful to distinguish several distinct types of "crashes": 1. those due to lack of backwards compatibility introduced by changes to the specification (e.g. 1738 to 2368 changes) 2. those due to faulty implementations of clear specifications 3. those due to implementations of different approaches due to ambiguity in the published specification In the case of #1, the horse has already escaped; we can certainly close the stable door by incorporating the 2822 restrictions on generation (conversely, we could make matters worse by introducing new incompatibilities such as raw untagged utf-8). In the case of #2, there will always be faulty implementations. There isn't much that can be done about that; non-conforming implementations are free to be such as long as conformance isn't claimed (and there is no enforcement anyway). Regarding #3, we can carefully review the specification for precision and clarity -- and that seems to be what's happening, though statements that something new is being introduced need to be carefully checked against what the draft under discussion and RFC 2368 actually say.
Received on Monday, 20 June 2005 19:25:49 UTC