** Observed with 4/20/99 Win32 build ** 1. Compose a message and add 2-3 addresses to "To:" and "CC:" lines. 2. Include one illegal address which contains 8-bit characters like this: grète@netscape.com The rest of the addresses should be legal ones. 3. Now send this msg. 4. This message will not go out to anyone. (A temporary file is created in the temporary directory, however.)
What do we want here? Do we want to generate address list with only valid addresses or includes invalid addresses? What if all the addresses are invalid, do we want to send the mail? Looks like this is not really i18n but more generic validation issue. Phil, any idea?
I think we should at least mimic 4.5x behavior. Let the server handle these kinds of errors. The client should NOT refuse to send out msgs simply because they contains illegal 8-bit characters in addr-spec.
So, the bug is that we don't send the message? If so, reassign to email@example.com. Note that we do have a preference for allowing 8bit headers, but it isn't visible in the UI. See the use of mail.strictly_mime_headers in msgsend.cpp cc'ing firstname.lastname@example.org to make sure I have this right.
I think we need a spec to deal with this kind of cases. Let me add info which has come to light since I filed this bug. 1. I sent 2-3 msgs containing 2-3 the following kind of addr-spec using Communicator 4.6: JPN@netscape.com along with at least one correct/legal address. This was several days ago. I still haven't received this msg at the legal address. However, I see this msg in my Sent folder on the IMAP server. I'm not sure if this msg really went out or not. Maybe it it still wandering somewhere. My current guess is that it did not. 2. If 1 is true, then 4.6 behavior is similar to what I reported for 5.0. 3. In either case, this is not desirable behavior. Based on 1 & 2, I'm changing my mind on this and want to suggest that we check for this kind of illegal addr-spec and ask the user to re-input in legal format before accepting to send it out. 4. By the way, looking at the source of the msgs stored in the Sent folder, I find that we B-encode the entire string including "@". See this example where the CC line contains the type of address I mentioned above. (Note: I still haven't received this msg at email@example.com.) Date: Thu, 29 Apr 1999 00:21:37 -0700 From: Katsuhiko Momoi <firstname.lastname@example.org> To: email@example.com CC: =?iso-2022-jp?B?GyRCRW0wZhsoQkBwb2x5Z2xvdC5tY29tLmNvbQ==?=
it must be a back-end issue. Reassigned to Rich
Will look into for M8.
Pushing out again :-( - rhp
This should work now. I've taken the approach that it will send the message and the valid addresses will get through and you will get a bounce message on the others. I don't like getting too fancy mucking with address information for fear of not sending something valid. I would rather error on the side of getting more bounce messages than not sending good email. Oh, in my testing, it seems that 4.5 would seemingly send the message, but nothing would ever get delivered and no bounce would happen so I think we are better off with 5.0 than where we were with 4.x. - rhp
** Checked with 8/13/99 Win32 build ** I'm going to mark this fix verified. I sent a message to 4 people with one of them containing 8-bit character in the address_spec itself. firstname.lastname@example.org email@example.com firstname.lastname@example.org "grète"@netscape.com I got the first 3 msgs but the 4th one was bounced with an error msg. This is in accord with the current fix. The only problem was that the 4th one got bounced for a wrong reason -- not because the user of the name "grète" does not exist but because the following recipient: Recipient: <" ="@netscape.com> Reason: <" ="@netscape.com>... User unknown does not exist. We are clearly doing something wrong when the "to: " line contains 8-bit characters either in NAME or real address part of: 8-bit_NAME <email@example.com>. See Bug 11892 for details.