Inline comments and mixing SUBS with types.

[Nick Mathewson <nickm@reputation.com>]

I've just checked in a more recent version of the spec.  To encourage
everybody to get it, I'm no longer mirroring it on wangafu.net.

I've modified George's convention of using [Q bracket-Q to indicate an
inline question] to include [Q initials of the person making the
comment.]

In general, though, I think we should get design issues out onto the
list, so here's one we ought to look at: Can SURB users set the message
type?  Otherwise, it would be quite hard e.g. to create SURBs for a
multipart message.  This would seem to return us to placing the type at
the head of the payload, rather than in the header: hmmm.

Here's what George and I wrote, for reference:

>[Q The few lines above describe how to decode a SURB, addressed
>to this mix. Would it be better if the type of message was
>actually specified by the real sender rather than the person that
>constructs the onion?
>
>WARNING: There might be some vulnerabilities in letting the
>sender specify the type of message. If any side effect resulting
>from the message being processed as a particular type is visible
>then this is threatening to the anonymity of the system. -GD]
>
>[Q Indeed.  Unless the type of the message is specifiable by the
>sender, how can anybody send a multipart message?  This needs to
>get straightened out. -NM] 

Yours,
-- 
Nick