Re: [Patch] Improve generation of obfuscated message-id

[Pawel Salek <pawsa0 gmail com>]

On 12/23/2016 04:29:11 PM, Albrecht Dreß wrote:
> Hi all,
>
> a while ago we introduced obfuscated message-id headers (see bugzilla  
> #738155 and related messages in the balsa list).
>
> The message-id now is basically a base64-encoded hash value, which  
> may contain the '+' and '/' characters explicitly allowed by RFC  
> 5322.  However, I noticed that if a DSN is requested, some MTA's seem  
> to use this message-id verbatim as ENVID parameter (see RFC 3461).   
> In this parameter, though, the '+' character is /not/ allowed and  
> must be escaped.  As a result, the receiving MTA /may/ reject the  
> message if the ENVID contains an unescaped +.
>
> Although this is a bug in the sending MTA, IMO it would be better to  
> adjust the way the message-id is created.  The attached patch uses  
> the first 240 bits (30 bytes) of the random hash to create a somewhat  
> formatted, base32-encoded message-id (see e.g. in this message).
>
> Opinions?

The goal seems sane. Have you got an idea how often the unwarranted  
rejection occurs?

Pawel