144901 - Comment node should not include comment delimiters

Status: RESOLVED WONTFIX

(Core :: DOM: HTML Parser, defect)

Description

[Heikki Toivonen (remove -bugzilla when emailing directly)]

If you have a comment like:

<!-- Foo -- -- bar -->

we currently create a DOM node with text: " Foo -- -- bar ". Arguably the text
should be: " Foo  bar ".

Comment 1

[Boris Zbarsky [:bzbarsky]]

We want to be able to serialize back out, though.... what about:

<!DOCTYPE -- this is text -- HTML PUBLIC ....>

Or even <!-- text -- FOO -- text -->

(I realize the second of these has some issues.....)

Comment 2

[Christopher Hoess (gone)]

From the DOM1 spec: Interface Comment 

     This represents the content of a comment, i.e., all the characters between 
the starting '<!--' and ending '-->'. Note that this is the definition of a 
comment in XML, and, in practice, HTML, although some HTML tools may implement 
the full SGML comment structure.  So I think, regrettably, that a "dumb" (i.e., 
non-SGML) interpretation is correct for comments within a comment declaration. 
As we currently parse HTML, we don't yet recognize comments in doctypes; I don't 
know what we do for XML (although even in DOM3, the documentType interface is 
fairly crude, and does not provide enough information to completely recreate an 
internal DTD subset from the DOM).

Ultimately, the DOM WG's decision to address HTML as if it were just an 
extension of the XML DOM Core, and not a SGML application, is at the root of 
these difficulties.

Comment 3

[Hixie (not reading bugmail)]

We should be compatible with Opera:
   http://weblog.timaltman.com/node/798#comment-869

The question is which browser should change. :-) Any ideas?

Comment 4

[Boris Zbarsky [:bzbarsky]]

As I said, any change to our code here would have to handle serialization properly...

Comment 5

[Sander]

Attachment: Testcase shows another comment problem

I was going to file a bugreport but this one seems to be sort of the same. Pieces of markup can drown between comment tags unwillingly. See the testcase.

Comment 6

[Boris Zbarsky [:bzbarsky]]

The behavior on the testcase in comment 5 is correct.  Sander, please go read up on SGML comment syntax and see the HTML spec's section on comments (and please don't comment on this bug with the results, since the whole thing has nothing to do with this bug).

Comment 7

[Stefan Thieme]

Attachment: More than Five Dashes in a Comment break Parser

Comment dashing Bug

SHORT:
if any of the html-comments contains more than five dashes, parser skips some of the subsequent html-code. if however the amount is less than six dashes the parsing works fine.

USAGE:

    * just open this html-document in your browser.
    * everything should be fine, watch the coloured page-source.
    * change the amount of dashes in a comment.
      i.e. increase them to contain a sequence of at least six dashes instead of five.
    * reload the document and see. 

VERIFIED:

    * Firefox 1.0.7 Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.12) Gecko/20051010 Firefox/1.0.7 (Ubuntu package 1.0.7)
    * Mozilla 1.8b2 Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8b2) Gecko/20050702
    * Minefield 3.0a1 Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a1) Gecko/20060524 Minefield/3.0a1
    * SeaMonkey 1.5a Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a1) Gecko/20060519 SeaMonkey/1.5a

Comment 8

[Boris Zbarsky [:bzbarsky]]

Stefan, see comment 6.  Your issue has nothing to do with this bug.

Comment 9

[warren crossing]

Look at what the w3c validator does if it encounters no DTD, it fallsback to HTML 4.01 Transitional. 

This bug persists regardless of doctype or xml declaration.  It can expose comments or hide large chunks of html between comments.  

I think if a doctype is given then gecko should respect the comment syntax of the xml.

Comment 10

[donelson]

Okay

Is this right? This bug has been know by Mozilla for THREE YEARS, and nothing has been fixed?

I wasted 3 hours today, trying to figure out why my website works fine on Webkit, IE, Opera, etc, but NOT on Firefox.

Why isnt this bug fixed?

Thanks
William

Comment 11

[Boris Zbarsky [:bzbarsky]]

Because it's not clear that it's a bug, or what the fix should be if it is.  Can you actually point to a spec that says what the behavior should be?

That said, I suspect you are not in fact seeing this bug; or are you really examining the content of Comment nodes in the DOM?

Comment 12

[donelson]

Boris, thanks.


When I do: View > Page Source...

on both Mac and PC Windows XP

<!-- Dummies ---
some javascript code here
-->

I see commented code which is green, and then in the middle of the commented code, the code suddenly becomes color-coded as active.

Removing the three dashes after "Dummies" fixes the bug in Firefox on my website.

So, yes, seems to fit with the bug description first reported over four years ago!

Comment 13

[Boris Zbarsky [:bzbarsky]]

Actually, that has nothing whatsoever to do with this bug.  What you're seeing there is that you have a comment which is malformed per HTML4 spec (unlike the comment described in comment 0 of this bug, which is just fine; the number of dashes matters), and that at this point Gecko is the only browser engine that actually tries to follow the HTML4 spec for comment parsing.  The proposed HTML5 draft changes how comments get parsed.  You might be interested in reading bug 214476 for the details.

In general, it's worth making sure you really understand the issues before shouting (which is what all caps typically means) about "three years" and so forth...

Comment 14

[donelson]

Thanks for your reply, and the reference.

Note:
None of the other browsers show this bad behaviour, regardless of the spec.

Text inside comments *should* behave in a predictable manner, and if it doesn't, it's sure to cause trouble. 

The "responsible" Standards committee members should behave rationally and considerately, not arrogantly and arbitrarily.

I've been programing for 40 years now, btw.

Comment 15

[Boris Zbarsky [:bzbarsky]]

Yes, but the issue in your case is the definition of "inside" comments.  "--" is the comment delimiter in HTML4.  Really, see bug 214476.  None of that has to do with this bug, though now the comments about it make up 1/3 of the comments here....

Comment 16

[Henri Sivonen (:hsivonen)]

WONTFIX per HTML5 spec.