接收邮件服务器在每个新行之前插入空格，打破multipart / alternative

Question

I am using PHP to send emails on demand to clients. I have a script which seemed fairly robust in testing, generating MIME-1.0 Compatible multipart/alternative emails that had a text and html version. Emails are sent as base64 encoded strings to preserve international characters (message text is usually in German).

However, it seems that certain servers, upon receiving the mail, insert a space (0x20) just before each CR-LF sequence. This doesn't break the base64, of course, but since it breaks up the CR-LF-CR-LF sequence that separates headers from messages, the messages are not parsed properly (or, at all, actually, since the secondary headers are never seen to stop).

Here is an example message as generated:

From: [e-mail]
To: [e-mail]
Subject: Test Message
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary="{$boundary}"

This is a multipart Message in MIME Format
--{$boundary}
MIME-Version: 1.0
Content-ID: <{$content_id}>
Content-Type: text/plain; charset="utf-8"
Content-Disposition: inline
Content-Transfer-Encoding: base64
Content-Length: {$objlen}

UkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVE
QUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNU
RUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQg
UkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVE
QUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNU
RUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQg
UkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVE
QUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNU
RUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQg
UkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVE
QUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNU
RUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQg
UkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVE
QUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNU
RUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQg
UkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQgUkVEQUNURUQ=
--{$boundary}
MIME-Version: 1.0
Content-ID: <{$content_id}>
Content-Type: text/html; charset="utf-8"
Content-Disposition: inline
Content-Transfer-Encoding: base64
Content-Length: {$objlen}

REVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVU
Q0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FE
RVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIg
REVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVU
Q0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FE
RVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIg
REVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVU
Q0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FE
RVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIg
REVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVU
Q0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FE
RVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIg
REVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVU
Q0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FE
RVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIg
REVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVIgREVUQ0FERVI=
--{$boundary}--

Is there some way to prevent the mail server from adding these spaces?

Answer 1

The problem has to do with certain email servers (e.g. t-online.de) treating CRLF newline sequences as less valid than LF only newlines. When newlines were changed from CRLF to LF, everything worked fine.

On the one hand, I would think this was a flagrant disregard for the standards set out in the RFCs, but on the other hand, I've had no issues with these messages since making the changes, so either (a) it doesn't matter or (b) there have been changes about which I do not know, which is always possible.

In any case, always end with LF only, I guess, if you intend to send multipart/* messages.