On Fri, 7 May 2004, Serge Munhoven wrote: > Hi, > > Inspiring from your trace, I hand-crafted requests and submitted them > with netcat6 [1] to host 64.233.167.104 on port 80. My diagnose would be > that there is a very weird buffer-overflow in Google's server with > regard to "Accept:" headers. Depending on the number of "Accept:" headers > and their size, the request fails (400) or succeeds (302 in my case: > redicrection to google.be). E.g., the following:
thanks - I was just starting to put together a test with LV's actual accept-string, to see what I could learn. > I dont understand exactly what overflows when. (Cf. a similar problem discussed > in August 2002 on this list: e.g, http://www.sysinternals.com server returned > a 406 when the "Accept:"-headers totalized more than 256 characters - but this > was an IIS; isn't google supposed to use Linux ? :). That was the theory. Maybe they're making it IIS-compatible (bug-for-bugs). > Conclusion: I don't really see how lynx should handle this differently. > Maybe check how other browsers handle a lot of mimetypes ? I seem to recall some observation that they don't advertise long lists of mimetypes. (It would be nice if other browsers had nice trace-files so one could answer these questions simply - though there's ethereal). w3m has a configure-screen text field which can be edited. That seems clumsy (if one does have a lot of mimetypes). A list of checkboxes seems to be overkill. -- Thomas E. Dickey http://invisible-island.net ftp://invisible-island.net _______________________________________________ Lynx-dev mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/lynx-dev
