On Thu, 2007-12-06 at 14:51 +0100, Peter J. Holzer wrote: > > 1. Don't do anything until there is an SMTP RFC which covers IDN. > > You misunderstood what I wanted a proposal for. I did not want a > proposal for IDN. I wanted one for a generalization of the parser so > that plugins can be used to change the acceptable syntax.
Not really. I object to both things ;-) However, I will now read the rest of your explanation on the second issue. I doubt that I object enough to make it worth another post. IDN is much worse. Please see my follow-up under the topic eai[*]. Work on SMTP is barely getting started but it appears that IDN and RFC 3490 are a collossal mistake (imo, not at all h) and should be avoided except where necessary. I think it would be far better to start with the new draft RFCs then to blindly implement RFC 3490. There seems to be an ESMTP extension (possibly only proposed) for allowing UTF-8 but this may just be for headers. This is referenced in the new eai drafts. I don't know whether RFC 3490 must be used in the envelope. Probably. That will be extremely annoying as you will have to translate addresses to compare. I believe that the ACE (see below) adopted was probably DUDE (which provides a one-to-one mapping to UTF-8, iirc -- some of the other encodings proposed did not, i think). That should be clear from reading 3490. Bernstein demonstrated that in 2001 UTF-8 already worked in the envelope for djbdns and qmail transparently and that trivial patches were required for bind and sendmail. Microsoft applications were already able to use UTF-8. Only after RFC 3490 was adopted, did microsoft (and some others) change their apps. So everything except DNS follows RFC2277 (i may have the number wrong -- from memory it's all 2s and 7s), which requires UTF-8 on the wire. Hopefully RFC 3490 has some allowance for a transition to UTF-8 ... some of the IDN WG suggested that replacing DNS was a better idea. See below for further amusement. [*] Subject: Email Address Internationalization --gh Summary of the IDN WG email list. --------------------------------- First of all, a few people on the IDN working group, including the chair, insisted that an ACE solution (ASCII Character-set Encoding -- or something like that) was absolutely required because of compatibility with existing 7-bit applications. In particular BIND and SENDMAIL. Bernstein argued that there were existing trivial patches or workarounds for BIND and SENDMAIL which would suffice. He proposed mandating these fixes with a reasonable time period allowed for transition. The difference between these two stands is that the first avoids short term pain for many sysadmins while the second avoids long term pain for all programmers writing network apps (perhaps a smaller group than the first). IETF has a tradition of taking the first path and Bernstein is quite vehement regarding the long term adverse consequences. He also pointed to RFC 2277, which required all future protocols to support UTF-8 on the wire. As usual, Bernstein was accused of ad-hominem attacks ... I don't agree with this assessment but I do agree that he is not particularly diplomatic. He did lose it completely at one point and used all caps for 3 sentences. He asked several times for a vote that there was a concensus that UTF-8 would be adopted eventually but his opponents claimed that he must write an RFC proposal first and he never did that, as far as I can tell. I don't know if this is considered in 3490 ... it should, at least, reference 2277 and justify non-adherence ... but I don't know until I read it and I may not bother.
