Re: lynx-dev 2.8.2dev.14 patch 2 - rules

[Klaus Weide]

On Mon, 25 Jan 1999, Leonid Pauzner wrote:

> > + # Examples:
> > + #   RULESFILE:/etc/lynx/cernrules
> > + #   RULE:Fail       file://localhost/etc/passwd* # Don't rely on it!
> > + #   RULE:Fail       gopher:*                     # reject by scheme
> > + #   RULE:Pass       finger://address@hidden/        # allow this,
> > + #   RULE:Fail       finger:*                     # but not others
> > + #   RULE:Map        http://old.server/*     http://new.server/*
> 
> Does '*' wildcard can be a string termination only,
> or probably like  http://*.com/*  etc.?

As far as I understand you can have 0 or 1 '*' in the first
URL, and it matches 0 or more characters.

If there is a second URL, and it also has a '*', then the '*'
gets replaced by the matches characters from the first URL.

Restricting URLs this way is in general not reliable, since there are
many ways to write the same thing - URLs don't get normalized to a
canonical form before the rules are checked.  For example one could
just append "#something" (I haven't checked whether that gets removed
before the rules are applied, I think not).  So at a minimum I would
end a URL with a '*' if there isn't already one.

Well that doesn't buy you much, the user could still hex-escape some
characters or use a double-'/', which makes the /etc/passwd example a
bad one.  The <scheme>:* forms should be reliable though (no
guarantees), matching is case-insensitive but is_url() will have
canonicalized the scheme (if it's a known one, otherwise we shouldn't
get this far).

    Klaus