- From: Dirk Herr-Hoyman <hoymand@gate.net>
- Date: Wed, 25 Jan 1995 19:52:25 -0500
- To: "Roy T. Fielding" <fielding@avron.ICS.UCI.EDU>, uri@bunyip.com
At 10:42 PM 1/25/95, Roy T. Fielding wrote: >> Headers are given in RFC822 format, and it is likely (and probably >> preferable) that only a "Subject:" header be included. Headers are given >> without spaces after them, such as "Subject:current-issue". > >I'd like to reiterate my opposition to this format. If the purpose of >this URL is to allow the retrieval of or subscription to resources of >a Mail server, there is no reason to allow any header fields to be >specified other than Subject. It is very unlikely for a mail server to >require special header fields as part of a request -- to do so would >remove access from all those users without knowledge and/or access to >rfc822 headers. In all cases where such header fields are used, the >mail server can allow them to be given at the beginning of the message >body, where they are relatively safe from transport effects. > >Arbitrary header fields introduce an unacceptable security hazard. There >is simply no way for a program to know, in advance, what header fields on >any given mail system could cause unintentional or undesired side-effects. >Having the user verify them does not significantly reduce the hazard -- >most users of these systems do not know what headers are, let alone >what they mean. Having the system ignore any unknown fields is equivalent >to not allowing them in the first place. > Although I don't see this as posing a threat, neither do a see any mail servers out that that depend on non-Subect header fields. Since I originally started the specification for the mailserver URL with the goal of providing a means to existing mail servers and extensibility second, I can live to the removal of arbitrary fields as a way of reaching consensus. >If we want to have a mailserver URL, the format should be > > mailserver:address[/<subject>[/<body>]] > Hmmm. This doesn't seem quite right to me. You are requiring that a Subject: field be included if there is a body. Many mail servers ignore the Subject: altogether. I was thinking more along the lines of using a //, i.e. a blank line, to signify the start of the body. mailserver:address/[<subject>/][/<body>] would work, for example. > >In addition, both subject and body should be restricted to disallow >the hex codes for CR and LF (%0D and %0A). Right, that we are letting the / serve as the end of line. Does make for a special case in the URL parsing, which I was hoping we could avoid, but it's not a terribly difficult one. > Also, the security section >should contain a warning to user agent developers to be aware of the >security limitations of an external mail agent if one is used to handle >the message. In particular, some mail agents allow special characters >to act as shell-escapes (to allow execution of commands within the mail >environment) and these need to be avoided by the user agent before >attempting to use the URL. > So, do we know what these shell escapes are? I'm assuming you are refering to something that causes a problem on the server side. If so, I'm not sure how we can come up with a list of prohibited characters. Or perhaps I'm just not comprehending the risk here, in which case could you elaborate? -- Dirk Herr-Hoyman <hoymand@gate.net> | I tried to contain myself CyberBeach Publishing | but * Internet publishing services | I got out Lake Worth, Florida, USA | Web: http://www.gate.net/cyberbeach/ Phone: +1.407.540.8309
Received on Wednesday, 25 January 1995 19:53:35 UTC