If the problem is with the implementation, we fix the implementation, not the protocol. We have similar arguments before.
-James Seng ----- Original Message ----- From: "Soobok Lee" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, May 29, 2002 1:14 AM Subject: Re: [idn] Re: Legacy charset conversion in draft-ietf-idn-idna-08.txt (in ksc5601-1987) > The big problem lies in that such bad and loose versioning practice is *everywhere*. > And precise and correct legacy-code versioning requires that "code-range&version" checking routines > should be inserted in every application. And such checking routines themselves also should be > versioned because legacy encodings are ever evolving. Both objectives requires upgrading all existing > i18n applications, but that may be not feasible, by the same reason why IDNA is preferred over UTF8 approach > in this WG. > > For example, let's assume future Outlook Express 7.0 will fix the bug by introducing > new mime charset name "KS_C_5601-1992" . The sender posts an email message in KS_C_5601-1992 > to the recipient who uses Outlook Express 6.0 which knows only "KS_C_5601-1987". > What will happen? maybe, fallback to iso8859-1 or default locale. > direct-charset-negotiation between the sender/recipient's email clients are not possible. > > A XML application server receives a XML request encoded in new KS_C_5601_1992, but the server applications > don't know the new charset. what happens if the transaction was in batch mode? There may be no immediate/interactive error report to > the orginator. > > That explains how much difficult it would be to introduce new version or new legacy encoding into > the IDN repertoires of supported encodings , if a certain version of IDN/IRI would have been widely deployed. > > Soobok Lee > > ----- Original Message ----- > From: "Roozbeh Pournader" <[EMAIL PROTECTED]> > > > On Wed, 29 May 2002, Soobok Lee wrote: > > > > > you can find the errornous mime-charset name : "KS_C_5601-1987". > > > Stupid and Wrong Versioning! > > > > Sure. But no protocol can fix broken software. Nag to developers of the > > software to fix it. In this case, it is passing a character onto wire, > > which is not in the character set it is claiming it to be in. > > > > If a piece of software wants to work with KSC 5601:1992, and use the > > character you used, should know its mapping to Unicode. Simple. > > > > roozbeh > > > > > > > > >
