164685 - Don't add ="" to custom boolean attributes

Status: NEW

(Core :: DOM: Serializers, defect)

Description

[Chris Weaver]

We use Maxum's NetCloak Pro [ http://www.maxum.com/NetCloak/ ] on our webserver
(G4 tower, OS X Server Unlimited running Apache).
To call up a server-side macro from NetCloak's macro file, one uses a <macro>
tag as such:
	<macro stylesheet>

In most other editors I use or have tried (DreamWeaver, PageSpinner, GoLive,
Freeway) these <macro> tags are ignored, since they are interpreted server-side.
In Mozilla's Composer, however, these tags (from a standard template we
distribute widely) are converted as below:
	<macro stylesheet=""></macro>

Obviously, NetCloak doesn't like this syntax and fails to use the <macro>
correctly. 
Is there any way to turn off Mozilla's syntax checking/rearrangement, per tag?
Thanks.

Comment 1

[Greg K.]

Confirmed using FizzillaCFM/1.1-final on 10.1.5. Reassigning to Editor: Core.

Comment 2

[Greg K.]

Attachment: Testcase

Composer adds ="" to the stylesheet boolean attribute in this testcase.

Comment 3

[kinmoz]

glazman, are we doing this internally? Or is this happening in the serializer?

Comment 4

[Daniel Glazman (:glazou) (not active in Mozilla any more)]

The parser handles boolean attributes in shorthand form only in the html
namespace for the time being. Fixing this bug would require adding to the parser
the ability to declare an attribute as boolean whatever is namespace, and the
implementation boolean attributes. Since boolean attributes in shorthand form
don't exist in XML, and since user-defined elements/attributes are not supposed
to be in a document at browsing time, I am not sure we're going to fix this...

Kin, what do you think ?

Comment 5

[Kathleen :Brade]

I thought the appropriate thing to output for booleans was something like
foo="foo"; this sounds like a duplicate bug but perhaps I misunderstand?

Comment 6

[BugBot [:suhaib / :marco]]

The bug assignee is inactive on Bugzilla, so the assignee is being reset.

Comment 7

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

I think that it's no problem to add ="", but I don't find any standards about serialization of boolean attributes. Anyway, this must be handled by serializer's staff.

Comment 8

[Nicolas CARPI]

Hello everyone,

I think that it's no problem to add =""

It is in fact a problem in some cases. For instance, this code:

<option selected>blah</option>

Will then get modified, when saving the page to disk, to:

<option selected="selected">blah</option>

However, if you're using JavaScript to save the HTML document (instead of using Ctrl+S), it might end up as selected="".

But if you're trying to run an HTML validator such as html-validate, with attribute-boolean-style rule, you'll get an error because selected is a boolean attribute and shouldn't have an empty string as value (in the context of the rule you chose to have).

To be precise, the spec allows for these 3 variations (naked, empty string, or attribute name). The problem here is the difference between what the server is actually sending / code shown in View source code (correctly shown without ="") and the code found after saving the html file on disk, which is different depending on how it was saved, and in no case similar to the "real" HTML sent by the server.

I understand this bug report was opened 22 years ago. Since then, spec has evolved, and I wish to propose that boolean attributes are allowed to stay "naked" in all cases. The issue here being about the browser allowing the user to access the exact same code that was sent by the server.

For reference: https://html.spec.whatwg.org/#boolean-attributes