Re: refuse-reject SMTP multi-lines replies

[Alexey Melnikov <[e-mail]>]

Дилян Палаузов wrote:

> Hi all,

Hi Дилян,

> Section 2.1.2 of draft-ietf-sieve-refuse-reject-06 speaks about how to 
> not-accpept a message (Action ereject, Rejecting a message at the 
> SMTP/LMTP protocol level).
>
> It might happen, that the ereject's parameter is longer than 512 
> characters, or it can be multilined (e.g
>   ereject "First line
>   second line";
> ). The line terminators within the script are not necessary CR and LF, 
> it can be only CR or only LF. (at least cyrus' timsieved allows only LF).

RFC 5228 only allows for CRLF as end-of-line markers, so this document 
shouldn't care about bare CRs or bare LFs used in scripts.

> Do you think it is appropriate to specify in the draft, when it comes 
> to SMTP, one of:
>
>     - If the ejerect's parameter consists of several lines, that lines 
> are merged into one single line. This line is split in portions of 
> about 500 characters (the number comes from RFC 2821, Section 4.5.3.1, 
> "reply line"; subtracting from 512 the length of "250"/"550" code and 
> the lengtt of enhanced status codes), when possible no word is 
> hyphenated. Each portion of the so created string-list is returned on 
> a separate smtp line in a multi-line SMPT reply;
>
>     - (If the ereject's parameter consists of several lines, the line 
> separators are replaced with CRLF.?)

As per RFC 5228, this statement is not needed.

> The SMTP makes multi-line reply, giving every line of the 
> reason-parameter on as a separate line in the SMTP dialog. (this 
> leaves unhandled the case, when a single line is longer than 512);
>
>     - If the ereject's parameter is longer than 500 characters, it is 
> truncated to 500 characters.

Personally, I don't like the 3rd choice. I would prefer to do a hybrid 
of your proposal 1 and proposal 2: try to preserve existing CRLFs in the 
text, but insert additional ones, if any single line is longer than 500 
characters.

>     Of course with the usual sense of "Shall/Must/Might/" etc. My 
> question is more if the draft shall deal with very-long-multi-lined 
> ereject-parameters, than what to do exactly.