| Message ID | 20220307033723.175553-3-jason@jasonyundt.email (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | gitweb: remove invalid http-equiv="content-type" | expand |
On Sun, Mar 06 2022, Jason Yundt wrote: > Before this change, gitweb would generate pages which included: > > <meta http-equiv="content-type" content="application/xhtml+xml; charset=utf-8"/> > > A meta element with http-equiv="content-type" is said to be in the > "Encoding declaration state". According to the HTML Standard, > > The Encoding declaration state may be used in HTML documents, > but elements with an http-equiv attribute in that state must not > be used in XML documents. > > Source: <https://html.spec.whatwg.org/multipage/semantics.html#attr-meta-http-equiv-content-type> > > This change removes that meta element since gitweb always generates XML > documents. > > Signed-off-by: Jason Yundt <jason@jasonyundt.email> > --- > gitweb/gitweb.perl | 4 +--- > t/t9502-gitweb-standalone-parse-output.sh | 13 +++++++++++++ > 2 files changed, 14 insertions(+), 3 deletions(-) > > diff --git a/gitweb/gitweb.perl b/gitweb/gitweb.perl > index fbd1c20a23..606b50104c 100755 > --- a/gitweb/gitweb.perl > +++ b/gitweb/gitweb.perl > @@ -4213,8 +4213,7 @@ sub git_header_html { > my %opts = @_; > > my $title = get_page_title(); > - my $content_type = get_content_type_html(); > - print $cgi->header(-type=>$content_type, -charset => 'utf-8', > + print $cgi->header(-type=>get_content_type_html(), -charset => 'utf-8', I think it would be better to just skip this hunk, no behavior will change if it's left in. > -status=> $status, -expires => $expires) > unless ($opts{'-no_http_header'}); > my $mod_perl_version = $ENV{'MOD_PERL'} ? " $ENV{'MOD_PERL'}" : ''; > @@ -4225,7 +4224,6 @@ 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"/> ..with this being the only behavior change (yeah the variable will now be used only in one place, but that's fine) I'm not sure I understand this change really. The result in always XML, so application/xhtml+xml is redundant, text/html, or both? But aside from that: I have seen browsers get the lack of encoding="" "wrong" with data at rest, don't some still default to ISO-8859-1? So won't this result in badly decoded data if you save the web page & view it locally? > <meta name="generator" content="gitweb/$version git/$git_version$mod_perl_version"/> > <meta name="robots" content="index, nofollow"/> > <title>$title</title> > diff --git a/t/t9502-gitweb-standalone-parse-output.sh b/t/t9502-gitweb-standalone-parse-output.sh > index e7363511dd..25165edacc 100755 > --- a/t/t9502-gitweb-standalone-parse-output.sh > +++ b/t/t9502-gitweb-standalone-parse-output.sh > @@ -207,4 +207,17 @@ test_expect_success 'xss checks' ' > xss "" "$TAG+" > ' > > +no_http_equiv_content_type() { > + gitweb_run "$@" && > + ! grep -Ei "http-equiv=['\"]?content-type" gitweb.body Nit: Should we skip the "-i" here since we're testing our own output, and not http standards in general (i.e. we don't have to worry about the case of http-equiv?)
On Monday, March 7, 2022 7:23:49 AM EST Ævar Arnfjörð Bjarmason wrote: > I'm not sure I understand this change really. The result in always XML, > so application/xhtml+xml is redundant, text/html, or both? To be honest, using an http-equiv="content-type" in XHTML is confusing. When you do use one, your goal shouldn’t really be to specify the document’s MIME type. After all, the first three lines of each page say <?xml version="1.0" encoding="utf-8"?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en-US" lang="en-US"> Those lines are more than enough to determine that something is using XHTML and UTF-8. Instead, the idea is to help out a parser that is incorrectly parsing the document as HTML (instead of as XHTML). Historical W3C documents (that were applicable when http-equiv="content-type" was allowed in XHTML) [1] [2][3] indicate that http-equiv="content-type" should be used like this: <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"/> In other words, to use http-equiv="content-type" properly in XHTML, you had to lie about the document’s type. The fact that this is confusing is probably part of why WHATWG disallowed it in the HTML Standard. > But aside from that: I have seen browsers get the lack of encoding="" > "wrong" with data at rest, don't some still default to ISO-8859-1? > > So won't this result in badly decoded data if you save the web page & > view it locally? I tested this idea in ungoogled-chromium, Firefox and Pale Moon. Other than Pale Moon in one specific circumstance, they all used UTF-8 as the encoding. Pale Moon used windows-1252, but only when the file ended with .html. When the file ended with .xhtml, Pale Moon used UTF-8. That being said, we don’t have to use an http-equiv="content-type" to fix the problem. Instead, we can use a <meta charset="utf-8"> which is allowed by the HTML Standard [4]. [1]: <https://www.w3.org/TR/xhtml1/#C_9> [2]: <https://www.w3.org/TR/html-polyglot/#character-encoding> [3]: <https://www.w3.org/Bugs/Public/show_bug.cgi?id=21818> [4]: <https://html.spec.whatwg.org/multipage/semantics.html#attr-meta-charset>
On 2022-03-07 at 03:37:23, Jason Yundt wrote: > Before this change, gitweb would generate pages which included: > > <meta http-equiv="content-type" content="application/xhtml+xml; charset=utf-8"/> > > A meta element with http-equiv="content-type" is said to be in the > "Encoding declaration state". According to the HTML Standard, > > The Encoding declaration state may be used in HTML documents, > but elements with an http-equiv attribute in that state must not > be used in XML documents. > > Source: <https://html.spec.whatwg.org/multipage/semantics.html#attr-meta-http-equiv-content-type> > > This change removes that meta element since gitweb always generates XML > documents. This change seems fine. We do specify this in the HTTP header, including the character set, which is what matters, so this should work in every browser, and the http-equiv is unneeded. I also don't think we need a meta header here, since we have an XML declaration, and that's controlling in this situation. This isn't regular HTML and we don't declare it as such, so using a meta header to control this isn't correct: the XML declaration should be used instead in the event a user downloads this to a local disk and processes it outside the context of an HTTP request. Since we control the HTTP headers, I'd actually argue that your test might well reject all http-equiv headers since they could be done much better with actual HTTP headers (and would therefore work with non-browser clients), but I don't think that's worth a reroll, nor do I think a test is even needed here (but bonus points for adding one). So I think this looks good as is. Thanks for the patch.
diff --git a/gitweb/gitweb.perl b/gitweb/gitweb.perl index fbd1c20a23..606b50104c 100755 --- a/gitweb/gitweb.perl +++ b/gitweb/gitweb.perl @@ -4213,8 +4213,7 @@ sub git_header_html { my %opts = @_; my $title = get_page_title(); - my $content_type = get_content_type_html(); - print $cgi->header(-type=>$content_type, -charset => 'utf-8', + print $cgi->header(-type=>get_content_type_html(), -charset => 'utf-8', -status=> $status, -expires => $expires) unless ($opts{'-no_http_header'}); my $mod_perl_version = $ENV{'MOD_PERL'} ? " $ENV{'MOD_PERL'}" : ''; @@ -4225,7 +4224,6 @@ 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 name="generator" content="gitweb/$version git/$git_version$mod_perl_version"/> <meta name="robots" content="index, nofollow"/> <title>$title</title> diff --git a/t/t9502-gitweb-standalone-parse-output.sh b/t/t9502-gitweb-standalone-parse-output.sh index e7363511dd..25165edacc 100755 --- a/t/t9502-gitweb-standalone-parse-output.sh +++ b/t/t9502-gitweb-standalone-parse-output.sh @@ -207,4 +207,17 @@ test_expect_success 'xss checks' ' xss "" "$TAG+" ' +no_http_equiv_content_type() { + gitweb_run "$@" && + ! grep -Ei "http-equiv=['\"]?content-type" gitweb.body +} + +# See: <https://html.spec.whatwg.org/dev/semantics.html#attr-meta-http-equiv-content-type> +test_expect_success 'no http-equiv="content-type" in XHTML' ' + no_http_equiv_content_type && + no_http_equiv_content_type "p=.git" && + no_http_equiv_content_type "p=.git;a=log" && + no_http_equiv_content_type "p=.git;a=tree" +' + test_done
Before this change, gitweb would generate pages which included: <meta http-equiv="content-type" content="application/xhtml+xml; charset=utf-8"/> A meta element with http-equiv="content-type" is said to be in the "Encoding declaration state". According to the HTML Standard, The Encoding declaration state may be used in HTML documents, but elements with an http-equiv attribute in that state must not be used in XML documents. Source: <https://html.spec.whatwg.org/multipage/semantics.html#attr-meta-http-equiv-content-type> This change removes that meta element since gitweb always generates XML documents. Signed-off-by: Jason Yundt <jason@jasonyundt.email> --- gitweb/gitweb.perl | 4 +--- t/t9502-gitweb-standalone-parse-output.sh | 13 +++++++++++++ 2 files changed, 14 insertions(+), 3 deletions(-)