Re: Annotations to CL-HTTP

To: schwarz@info02.uni-trier.de (Elmar Schwarz)
Subject: Re: Annotations to CL-HTTP
From: JCMA@ai.mit.edu (John C. Mallery)
Date: Fri, 15 Sep 1995 17:37:10 -0400
Cc: pch@mystech.com, www-cl@ai.mit.edu
Sender: owner-www-cl@ai.mit.edu
At 8:25 AM 9/15/95, Elmar Schwarz wrote:
>Hi,
>
Hi Elmar,

Can you tell us which version of CL-HTTP and what version of lisp
you are using?

>there occured some problems using 'redirect'ed URLs using mac-cl-http with
>an HTTP-call from an IBM compatible using netscape 1.1. Netscape just
>displayed and error message like 'Socket not found'. Using the 'true' URL
>with an attached response-function everything worked fine. That might be
>ineresting for you.

Are you saying that two levels of indirect do not work?  Please note
that redirects in the current specification must go to a client, not to a
server.
In http 1.1,  it will be possible redirect through URLs on the local server.
There is code in place to do this, but it won't be turned on in the releases
systems until there are 1.1 clients.

>
>Another thing is just an annotation to the HTML 3.0 (netscape.lisp)
>extensions. I've tried to use an background image which (in the
>specification that I do know) is available by an BACLGROUND=... option in
>the BODY tag. As I looked into the HTML-source created by your
>netscape::with-document-body function, I saw that the background
>key-argument was converted to a BGCOLOR option, resulting in funy
>colors. Well I'd find it use full to provide two key-parameters: BGCOLOR and
>BACKGROUND which should be translated one to one into HTML. This would make
>the handling more easy I guess. (Just a suggestion.)

The reason we have one argument is to reintegate the abstraction split by
Netscape.
Background goes to background= when a URL is provided and to bgcolor when a
color
keyword is supplied.  If you want to pass in a color code "#ffff", we cn
easily recognize
these passed in via background.  Is this what you want to do?

Or, are you trying to overlay a partially transparent image on a buit-in
color?  If this
latter case makes, sense then we'll have to split the abstraction back so
that you can
control these parameters independently.

>
>A last thing has to do with the old LINEFEED and CARRIAGE RETURN
>problem. I'm not absolutely sure, but I think it might be better to output
>both a CR and a linefeed for any freshline command. This seems to be the more
>convenient way, that has two advantages:
>
>-Some Browsers (Old Mosaic) don't handle a single CR as a line break in
> preformatted text resulting in an unreadable page.

Shouldn't people just FTP the new version?

>-Source produced by MAC-CL-HTTP is absolutely unreadable on machines like
> IBM compatibles or UNIX workstations, since anything is written into a
> single line.
>
>Well since I want to go to the whole world with our server and these two
>systems are more common than the Mac is, it might be a reasonable change.

There are a number of ways to do this.

It sounds like you want this only for generated html.  Would you be
satisfied with
a function that writes CR-LF to the http stream? Or are you asking for it
to be built
into to fresh-line in the html generation packages?