Re: lynx-dev Query string handling bug vs bad html

[Foteos Macrides]

Chris Gray <address@hidden> wrote:
>Doug,
>
>Fote and Alan are right, and Lynx is right.  Just so you have it three
>times. :)

        Well, except that when the majordomo processed my message, it
signalled that I may not be just typo prone, but getting dyslexic in
my old age. :)  That named character reference was "lg" (less than or
greater than), not "gl".

        This kind of thing is going to be an FAQ (IMHO), so it might
also be worthwhile spelling out the related issues.  It's true that
early implementations of SGML character reference handling in text/html
were very poor for the commercial GUIs, but the more recent versions do
seem to reflect recognition by their implementors that failure to SGML
escape ampersands which do not introduce a character reference is
equivalent to not URL (hex) escaping a percent sign in a URL when it
does not introduce hex notation for a high-value character in the URL.
However, some people who know what is "correct" to do might none-the-less
choose names for the name=value pairs in queries that are not in the
HTML specs, and not SGML escape the ampersand to accomodate those older
browsers.  All browsers use the same error recovery as Lynx in that
case, so it "should" be OK.  However, Lynx v2.8 went way, way beyond the
HTML 4.0 specs in its recognition of SGML named entities (i.e., it
recognizes many, many which are not in the HTML 4.0 specs).  The "lg"
is among those, and it's conceiveable that someone might use it without
SGML escaping the preceding ampersand, thinking that it should thus work
for both old and newer browsers, but in fact causing a failure for Lynx
v2.8 (that's one of the problems with the attitude that one need not
make oneself familiar with what other browsers have or have not done
when programming for Lynx).

                                Fote
-- 
Foteos Macrides (address@hidden during April, '98)