From: Carsten S. <ca...@gn...> - 2003-08-01 21:48:02
|
It seems I sent the following to Sean only, although it was meant for the list. I have committed the changes in the meantime. Hi Sean, thanks for the answer. > > However, I dislike the current solution and would like to replace > > it entirely, executing scripts and appmods but discarding any > > content they generate. This would make the headers being sent to > > the client more accurate and would allow following {page, ...} > > redirects, which I think is important. > > This sounds OK to me. Web designers are not supposed to use GET > requests if they want side effects so there should be no danger in > executing scripts etc. Fine. > The other option would be to just miss out the Content-Length field > from HEAD responses which would have generated dynamic content - I > think this is still kind of valid HTTP/1.1 It's not just the Content-Length field. I am more concerned about Etag and Last-Modified. I am having a script icons.yaws which more or less says <erl> out(A)-> FormatExt case (A#arg.headers)#headers.user_agent is a sensible Browser of true -> ".png"; false -> ".gif" end, {page, "/icons" ++ A#arg.pathinfo ++ FormatExt}. </erl> With this setup http://www.fu-mathe-team.de/icons.yaws/fumt-logo-ball-white will give you either a GIF (for Netscape 4.7x or, should you have access to such a system, MSIE on Windows) or a PNG. These have different entity tags and modification dates which are correctly transmitted in response to a GET request, but with the current version of yaws, these are not delivered in response to a HEAD request. This makes a difference for `wget -m' for example. I have implemented the proposed changes, but will have to check, if the connection is kept open or closed appropriately, before committing. Greetings, Carsten --=20 Carsten Schultz (2:40, 33:47), FB Mathematik, FU Berlin http://carsten.fu-mathe-team.de/ PGP/GPG key on the pgp.net key servers,=20 fingerprint on my home page. |