Re: New syntax on message-IDs

[Russ Allbery <rra@xxxxxxxxxxxx>]

"Charles Lindsey" <chl@xxxxxxxxxxxxxxxx> writes:

> OK, I got that now. It essentially means "I have an article with this
> msg-id; would you like me to send it?" But in fact there is no great
> need for the server to check the syntax of the msg-id (though there is
> nothing wrong or harmful in doing so). The client has no business
> offering a bad msg-id in the first place (and wherever it got it from
> should not have sent it, right back to who/whatever constructed it in
> the first place).

Yes, there's no major need, but it's not a bad idea anyway.  There is some
interoperability justification in rejecting malformed message IDs, since
some servers behave poorly if you send them one, in ways that can cause
problems in correctly identifying the error and understanding which
article to not send on to the badly behaving server.  Rejecting them
up-front is in some ways cleaner.

This is particularly true of excessively long message IDs, which can cause
all sorts of weird problems.

Message IDs containing whitespace are in general just a bad idea, too,
since they're likely to break NNTP command parsing in numerous places down
the line.

-- 
Russ Allbery (rra@xxxxxxxxxxxx)             <http://www.eyrie.org/~eagle/>