httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From r..@ai.mit.edu (Robert S. Thau)
Subject Re: Bugs in getparents()
Date Thu, 09 Nov 1995 20:56:37 GMT
Hmmm... at first blush, this looks OK (though I haven't tried running
with it yet).  Two comments:

1) The combination of this patch and my initial "fix" to the
   ScriptAlias non-bug would "fix" every print-the-script gotcha
   *that* *I'm* *currently* *aware* *of*, without breaking currently
   valid URIs or sending new redirects.  I could live with that, but
   I'm still not sure we're aware of all such anomalies, and would not
   be entirely comfortable telling the users (even implicitly, by not
   warning them about these sorts of problems) that we did.  Still, I
   could live with it, if it would help get something out the door.

2) It might be better to resolve /foo//../bar to /bar rather than
   /foo/bar.  That might not be *strictly* in accord with RFC 1808,
   but it's not much of a deviation, it would probably be more in
   accord with user expectations, and it would eliminate the
   <!--#include virtual="../foo"--> anomaly, not that I consider that
   any big deal either.  This would involve adding two extra lines of
   code to the rewritten getparents() routine, one each in parts three
   and four (loops skipping over extra slahes preceding '/..', viz. my
   earlier getparents() patch).

rst

Mime
View raw message