From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date: Tue, 08 Mar 2022 12:44:35 GMT
Subject: Re: [PATCH v2 2/2] gitweb: remove invalid http-equiv="content-type"
Message-ID: <220308.86cziwy595.gmgdl@evledraar.gmail.com>
In-Reply-To: <Yia2Y2zoOjvVdJ9O@camp.crustytoothpaste.net>


On Tue, Mar 08 2022, brian m. carlson wrote:

> [[PGP Signed Part:Undecided]]
> On 2022-03-08 at 01:07:11, Jason Yundt wrote:
>> diff --git a/gitweb/gitweb.perl b/gitweb/gitweb.perl
>> index fbd1c20a23..59457c1004 100755
>> --- a/gitweb/gitweb.perl
>> +++ b/gitweb/gitweb.perl
>> @@ -4225,7 +4225,7 @@ sub git_header_html {
>>  <!-- git web interface version $version, (C) 2005-2006, Kay Sievers <kay.sievers\@vrfy.org>, Christian Gierke -->
>>  <!-- git core binaries version $git_version -->
>>  <head>
>> -<meta http-equiv="content-type" content="$content_type; charset=utf-8"/>
>> +<meta charset="utf-8"/>
>
> I don't actually think this is an improvement.  I don't think it's
> necessary, considering we have an XML declaration and the HTTP header,
> both of which already say it's UTF-8 and will take precedence over this.

Ageed. I was a bit surprised per Jason's
https://lore.kernel.org/git/109813056.nniJfEyVGO@jason-desktop-linux/
that the removal wasn't kept.

I.e. he was replying to a question of mine asking whether we didn't need
this data at rest, e.g if you save the page. I didn't notice the "<?xml
version..." we emit, which seems to be enough.

I.e. this seems to have always been redundant going back to c994d620cc8
(v220, 2005-08-07), or rather, the character set part of it.

Maybe I still don't understand this, but the commit message seems to me
be conflating whether we send the *right* http-equiv with whether we
send it at all, i.e. if the problem is that XML documents shouldn't be
text/html isn't this correct?:
	
	diff --git a/gitweb/gitweb.perl b/gitweb/gitweb.perl
	index fbd1c20a232..c1c5af0b197 100755
	--- a/gitweb/gitweb.perl
	+++ b/gitweb/gitweb.perl
	@@ -4049,7 +4049,13 @@ sub get_page_title {
	 	return $title;
	 }
	 
	+sub get_content_type_xml {
	+	return 'application/xhtml+xml';
	+}
	+
	 sub get_content_type_html {
	+	my ($want_xml) = @_;
	+
	 	# require explicit support from the UA if we are to send the page as
	 	# 'application/xhtml+xml', otherwise send it as plain old 'text/html'.
	 	# we have to do this because MSIE sometimes globs '*/*', pretending to
	@@ -4057,7 +4063,7 @@ sub get_content_type_html {
	 	if (defined $cgi->http('HTTP_ACCEPT') &&
	 	    $cgi->http('HTTP_ACCEPT') =~ m/(,|;|\s|^)application\/xhtml\+xml(,|;|\s|$)/ &&
	 	    $cgi->Accept('application/xhtml+xml') != 0) {
	-		return 'application/xhtml+xml';
	+		return get_content_type_html();
	 	} else {
	 		return 'text/html';
	 	}
	@@ -4214,6 +4220,7 @@ sub git_header_html {
	 
	 	my $title = get_page_title();
	 	my $content_type = get_content_type_html();
	+	my $content_type_xml = get_content_type_html();
	 	print $cgi->header(-type=>$content_type, -charset => 'utf-8',
	 	                   -status=> $status, -expires => $expires)
	 		unless ($opts{'-no_http_header'});
	@@ -4225,7 +4232,7 @@ sub git_header_html {
	 <!-- git web interface version $version, (C) 2005-2006, Kay Sievers <kay.sievers\@vrfy.org>, Christian Gierke -->
	 <!-- git core binaries version $git_version -->
	 <head>
	-<meta http-equiv="content-type" content="$content_type; charset=utf-8"/>
	+<meta http-equiv="content-type" content="$content_type_xml; charset=utf-8"/>
	 <meta name="generator" content="gitweb/$version git/$git_version$mod_perl_version"/>
	 <meta name="robots" content="index, nofollow"/>
	 <title>$title</title>

Of course we might then *also* decide that <meta http-equiv> in this
case isn't needed at all, but isn't that a seperate change?

And won't conforming browsers treat application/xhtml+xml differently
when the page is saved? A long time ago (Idid some web development)
using it would enable pedantic strictness in browsers, i.e. unclosed
tags etc. would be a hard error, but I can't reproduce that locally in
either Firefox or Chrome now (with just the gitweb output as-is with
that http-equiv tweaked).

So maybe it does nothing, or maybe it's just those browser...