Re: LYNX-DEV fotemods chartrans & SSL

[Klaus Weide]

On Thu, 23 Oct 1997, Foteos Macrides wrote:

> Klaus Weide <address@hidden> wrote:
> >>[...]
> >>    This was explained at length in the fruitless "discussions"
> >> before Klaus' extended disappearance, and again more recently.  I'm not
> >> optimistic that I can get it across any better than I've tried before,
> >> but I'll give it another shot.  I use Lynx seriously in conjunction
> >> with secure transactions.  In that case, it's caching and resubmission
> >> logic are very important.  The INTERNAL_LINK stuff breaks that logic.
> >> It may not resubmit form content when it should, which is not so bad
> >> because you can force resubmission in such cases, and you usually
> >> don't want to resubmit in most form-based secure transactions.  The
> >> problem is that it also will force resubmissions when it shouldn't,
> >> which is very bad in such cases, and precludes my using that code
> >> seriously.
> >
> >Please give a concrete example.
> 
>       I gave plenty of concrete examples, plus demonstrations, in our
> "discussions" about this, and no, I don't plan to get into another round
> with you about it.

Giving a URL doesn't mean you have to get into either a discussion or
"discussion".

I know about
        <URL:http://www.wfbr.edu/lynx/no-cache.html>
        <URL:http://www.wfbr.edu/lynx/expires.html>,
I am looking for an indication that any existing service is relying on a
specific behavior, one way or another.

>       The only new information that might be worth posting is that
> before submitting the latest revision Roy Fielding contacted me for
> feedback and we discussed whether it was still subject to the
> misinterpretations that crept in when Roy was in Australia and Larry
> Masinter tried to do a revision on his own.  I replied that it was fine
> now, but the revision has been out for some time, the INTERNAL_LINK
> stuff is still in the devel code, so perhaps I gave Roy bad feedback.
> 
>       One other thing that I didn't include in my reply to
> Henry.  The revision now makes clear that a form's ACTION always
> is resolved versus the base, rather than (current document), even
> if the ACTION is empty.  In the devel code, you not only have the
> problem of faulty caching and resubmission logic, but could end
> up trying to submit the form to the wrong address.   Also very
> bad if it's a secure transaction.

I will look at that.

In any case, I believe that no service can rely on specific behavior
for this, since it hasn't been implemented the same way in clients,
and specs (or drafts) have been changing.  An author who would expect
a specific behavior for important or sensitive data would be foolish.

> >>[...]
> >> <!--LynxCSI-->
> >
> >For those who want to try it, make that 
> >  <!--#LynxCSI-->
> >[...]
> >Yes, that change was made in 2.7.1ac-0.61.  The "example interface"
> >broke when I added the -source -preparsed stuff (lynx -preparsed
> >crashed on www.microsoft.com which has comments that start with
> >"<!--#"). Therefore I had to change something.  I made the
> >modification that seemed most obvious.
> >[...]
> >(There _was_ a serious error in 2.7.1ac-0.83, where I had already
> >changed back the interface in SGML.c but had forgotten to make the
> >equivalent change in LYCharUtils.c.  And yes I should have tested that
> >but didn't.  It was corected in 2.7.1ac-0.84.  That doesn't seem to
> >be what you are talking about though.)
> >[...]
> >I honestly think that you are completely wrong, and that you have not
> >checked your statements.  It seems I haven't found a way to disagree
> >with you that you wouldn't take personally.
> 
>       Yes, I'm typo prone and left off the #, and, yes, I was
> misremembering -83 as -84, and, no, I don't know what bug in -63
> when you were working on the -preparsed stuff might have become
> manifest with a crash there for www.microsoft.com.  However, it
> would always be called with the same target argument as when
> HTML_start_element() or HTML_end_element() are called, which do
> not check the isa signature, www.microsoft.com still has the
> <!-# comment, and if you remove the isa signature check from
> LYCheckForCSI(), it will not crash when accessing that, with
> or without -source and/or -preparsed.   So your interdigitated
> comments still seem inappropriate.

But HTML_start_element() and HTML_end_element() are not any more
the only functions that may be called by SGML.c's start_element()
and end_element().

The following is with 2.7.1ac-0.84 (-87 should be the same):

sol:lynx2-7-1aclatest$ gdb lynx
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.16 (sparc-sun-solaris2.4),
Copyright 1996 Free Software Foundation, Inc...
(gdb) break LYCheckForCSI
Breakpoint 1 at 0x79818: file ./LYCharUtils.c, line 4467.
(gdb) set args http://www.microsoft.com
(gdb) run
 [ ... ]
Read 1460 bytes of data.
Breakpoint 1, LYCheckForCSI (me=0x14cc78, url=0x14acdc) at ./LYCharUtils.c:4467
4467        if (!me)
(gdb) list
4462    */
4463    PUBLIC BOOLEAN LYCheckForCSI ARGS2(
4464            HTStructured *,         me,
4465            char **,                url)
4466    {
4467        if (!me)
4468            return FALSE;
4469
4470        if (!me->isa || strcmp(me->isa->name, "Lynx_HTML_Handler"))
4471            return FALSE;
(gdb) print *me
$1 = {isa = 0xd2660, node_anchor = 0x1495b0, text = 0x14ad78, target = 0x0,
  targetClass = {name = 0x0, _free = 0, _abort = 0, put_character = 0,
    put_string = 0, put_block = 0}, CurrentA = 0x1421f8, CurrentANum = 0,
 [ ... ]
(gdb) print me->isa
$2 = (HTStructuredClass *) 0xd2660
(gdb) print *me->isa
$3 = {name = 0xd2688 "Lynx_HTML_Handler", _free = 0x67630 <HTML_free>,
  _abort = 0x679f4 <HTML_abort>, put_character = 0x592ec <HTML_put_character>,
  put_string = 0x59868 <HTML_put_string>, _write = 0x59ca0 <HTML_write>,
  start_element = 0x59ce4 <HTML_start_element>,
  end_element = 0x64b58 <HTML_end_element>,
  put_entity = 0x6755c <HTML_put_entity>}
(gdb) kill
Kill the program being debugged? (y or n) y
(gdb) set args -source -preparsed http://www.microsoft.com
(gdb) run
 [ ... ]
Breakpoint 1, LYCheckForCSI (me=0x141900, url=0x141b9c) at ./LYCharUtils.c:4467
4467        if (!me)
(gdb) p *me
$4 = {isa = 0xe689c, node_anchor = 0x1417a0, text = 0xd28d8,
  target = 0x687e4, targetClass = {name = 0x68e48 "\235c?\220!",
    _free = 0x68764 <HTFWriter_put_character>,
    _abort = 0x687ac <HTFWriter_put_string>,
    put_character = 0x687c4 <HTFWriter_write>,
    put_string = 0x3c2f5441 <MessageSecs+1008458489>,
    put_block = 0x424c453e <MessageSecs+1111018486>}, CurrentA = 0xd0a6172,
  CurrentANum = 2033150317,
 [ ... ]
(gdb) p me->isa
$5 = (HTStructuredClass *) 0xe689c
(gdb) p *me->isa
$6 = {name = 0xe68c0 "HTMLGen", _free = 0x983e0 <HTMLGen_free>,
  _abort = 0x98448 <HTMLGen_abort>,
  put_character = 0x97cb8 <HTMLGen_put_character>,
  put_string = 0x97ffc <HTMLGen_put_string>, _write = 0x98034 <HTMLGen_write>,
  start_element = 0x9806c <HTMLGen_start_element>,
  end_element = 0x982cc <HTMLGen_end_element>,
  put_entity = 0x98384 <HTMLGen_put_entity>}
(gdb)



     Klaus

;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;