Re: LYNX-DEV HTML Form Test

[Foteos Macrides]

address@hidden (Wilson Cheung) wrote:
>This was mentioned a while ago in one of the recent messages about Cookie 
>support and such but I wasn't sure if anyone had resolved to get things 
>"fixed" or to ignore it (I must have deleted one too many messages the
>last time I was pruning my email).
>
>There is a HTML Form Test page at:
>
>http://www.research.digital.com/nsl/formtest/home.html
>
>
>According to its list of stats, the current version of Lynx 2.6FM 01/18/96
>still "fails" two tests:
>
>NetscapeCookie       3  37%  Bad HTTP Cookie value: COOKIE=testvalue;
> $Path=/nsl/formtest; $Domain=www.research.digital.com
>[...]
>NewlineChar          7 100%  Browser uses LF in <TEXTAREA>
>[...]
>
>Their definition of a Cookie seems to be based on the old specifications 
>provided by Netscape Corp., though.  It might not be an accurate test for 
>the current implementation which is based on the recent RFC draft Fote listed.

        It's hard to say whether the cookie test is misimplemented by
the formtest server, or whether the instructions on how to make the IETF
version of cookies backward compatible with Netscape's "historical"
implementation are lies.  Lou Montulli (who went to Netscape on graduating
from UKans in 1994, and is still at Netscape) is a secondary author of
the IETF draft, so I have been assuming that the backward compatibility
instructions are not lies.

        Anyway, the formtest server is expecting only:
        
        COOKIE=testvalue

and is objecting to the:

         ; $Path=/nsl/formtest; $Domain=www.research.digital.com

However, the IEFT draft says the using "; $foo=blah" for the extensions,
instead of just "; foo=blah" (i.e., '$' prefixes for each name in the
extended name=value pairs), makes them backward compatible for Netscape.
Sigh...



        The newline message is just a "warning".  Here's the story on
that.  Lou implemented the Lynx FORM support in 1993 using only a
hex escaped LF (%0a) as the newline character in TEXTAREAs, like
XMosaic v2.0, which had just "invented" HTML FORMs.

        Years latter, the IETF RFC for HTML 2.0 specified that the
newlines must be hex escaped CRLF (%0d%0a), so I changed Lynx
accordingly, and got gobs of "bug reports" about Lynx adding '^M's
to all the lines (from Unix CGI script writers, of course 8-).
So I changed it back to just LF (%0a) and there have been no
"bug reports" about that since.

        Then the IETF HTML 3.0 specs said that "both servers and
clients should be tolerant of CR, LF, or CRLF as newlines" in
TEXTAREAs, so I modified Lynx to accept all three as newlines,
but still use just LF (%0a) when sending FORM content that is
application/x-www-urlencoded.  The W3C HTML 3.2 Recommendation
still just "recommends" CRLF, but allows CR or LF.

        It's a simple mod in HText_SubmitForm() of GridText.c
if you want to try the CRLF "recommendation" and see if it
still trips up CGI scripts on Unix servers.  If the ACTION
is a mailto URL, most mail software will handle both LF and
CRLF as newlines, but there is still some around (written
is C) that will display '^M's on the ends of lines in the 
message body (that's probably another "what terminal emulator
are you using, and what's in your termcap or terminfo"
pickles 8-).

                                Fote

=========================================================================
 Foteos Macrides            Worcester Foundation for Biomedical Research
 address@hidden         222 Maple Avenue, Shrewsbury, MA 01545
=========================================================================
;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;